0% found this document useful (0 votes)

106 views

Open CVReference Manual

Uploaded by

Ayush Maheshwari

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

106 views

Open CVReference Manual

Uploaded by

Ayush Maheshwari

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 417

Open Source Computer Vision Library

Reference Manual

Version -001

Version History Original Issue

Date December 8, 2000

This OpenCV Reference Manual as well as the software described in it is furnished under license and may only be used or copied in accordance with the terms of the license. The information in this manual is furnished for informational use only, is subject to change without notice, and should not be construed as a commitment by Intel Corporation. Intel Corporation assumes no responsibility or liability for any errors or inaccuracies that may appear in this document or any software that may be provided in association with this document. Except as permitted by such license, no part of this document may be reproduced, stored in a retrieval system, or transmitted in any form or by any means without the express written consent of Intel Corporation. Information in this document is provided in connection with Intel products. No license, express or implied, by estoppel or otherwise, to any intellectual property rights is granted by this document. Except as provided in Intel's Terms and Conditions of Sale for such products, Intel assumes no liability whatsoever, and Intel disclaims any express or implied warranty, relating to sale and/or use of Intel products including liability or warranties relating to fitness for a particular purpose, merchantability, or infringement of any patent, copyright or other intellectual property right. Intel products are not intended for use in medical, life saving, or life sustaining applications. Intel may make changes to specifications and product descriptions at any time, without notice. Designers must not rely on the absence or characteristics of any features or instructions marked "reserved" or "undefined." Intel reserves these for future definition and shall have no responsibility whatsoever for conflicts or incompatibilities arising from future changes to them. The OpenCV may contain design defects or errors known as errata which may cause the product to deviate from published specifications. Current characterized errata are available on request. Intel and Pentium are registered trademarks of Intel Corporation or its subsidiaries in the United States and other countries. *Other names and brands may be claimed as the property of others. Copyright Intel Corporation 2000-2001.

Eigen Objects ......................................................................................... 5-1 Embedded Hidden Markov Models ........................................................ 5-2

OpenCV Reference Manual

Contents

Chapter 6 3D Reconstruction
Camera Calibration................................................................................. 6-1 Camera Parameters ...................................................................... 6-1 Homography.................................................................................. 6-2 Pattern........................................................................................... 6-3 Lens Distortion .............................................................................. 6-4 Rotation Matrix and Rotation Vector ............................................. 6-5 View Morphing........................................................................................ 6-5 Algorithm ....................................................................................... 6-6 Using Functions for View Morphing Algorithm .............................. 6-8 POSIT..................................................................................................... 6-9 Geometric Image Formation ....................................................... 6-10 Pose Approximation Method ............................................................. 6-11 Algorithm ..................................................................................... 6-13 Gesture Recognition............................................................................. 6-15

Chapter 7 Basic Structures and Operations

Image Functions ..................................................................................... 7-1 Dynamic Data Structures........................................................................ 7-4 Memory Storage ................................................................................. 7-4 Sequences.......................................................................................... 7-5 Writing and Reading Sequences ........................................................ 7-6 Sets..................................................................................................... 7-8 Graphs .............................................................................................. 7-11 Matrix Operations ................................................................................. 7-15 Drawing Primitives ................................................................................ 7-15 Utility..................................................................................................... 7-16

Chapter 8

Library Technical Organization and System Functions

Error Handling ........................................................................................ 8-1 Memory Management ............................................................................ 8-1

OpenCV Reference Manual

Contents

Interaction With Low-Level Optimized Functions.................................... 8-1 User DLL Creation.................................................................................. 8-1

Chapter 9 Motion Analysis and Object Tracking Reference

Background Subtraction Functions......................................................... 9-3 Acc..................................................................................................... 9-3 SquareAcc ......................................................................................... 9-4 MultiplyAcc......................................................................................... 9-4 RunningAvg ....................................................................................... 9-5 Motion Templates Functions................................................................... 9-6 UpdateMotionHistory ......................................................................... 9-6 CalcMotionGradient ........................................................................... 9-6 CalcGlobalOrientation........................................................................ 9-7 SegmentMotion.................................................................................. 9-8 CamShift Functions ................................................................................ 9-9 CamShift ............................................................................................ 9-9 MeanShift......................................................................................... 9-10 Active Contours Function ..................................................................... 9-11 SnakeImage..................................................................................... 9-11 Optical Flow Functions ......................................................................... 9-12 CalcOpticalFlowHS .......................................................................... 9-12 CalcOpticalFlowLK........................................................................... 9-13 CalcOpticalFlowBM.......................................................................... 9-13 CalcOpticalFlowPyrLK ..................................................................... 9-14 Estimators Functions ............................................................................ 9-16 CreateKalman.................................................................................. 9-16 ReleaseKalman ............................................................................... 9-16 KalmanUpdateByTime..................................................................... 9-17 KalmanUpdateByMeasurement....................................................... 9-17 CreateConDensation ....................................................................... 9-17 ReleaseConDensation..................................................................... 9-18

OpenCV Reference Manual

Contents

ConDensInitSampleSet ................................................................... 9-18 ConDensUpdatebyTime................................................................... 9-19 Estimators Data Types.......................................................................... 9-19

Chapter 10 Image Analysis Reference

.............................................................................................................. 10-1 Contour Retrieving Functions ............................................................... 10-6 FindContours ................................................................................... 10-6 StartFindContours............................................................................ 10-7 FindNextContour.............................................................................. 10-8 SubstituteContour ............................................................................ 10-9 EndFindContours ............................................................................. 10-9 Features Functions............................................................................. 10-10 Fixed Filters Functions.................................................................... 10-10 Laplace .......................................................................................... 10-10 Sobel.............................................................................................. 10-10 Feature Detection Functions........................................................... 10-11 Canny............................................................................................. 10-11 PreCornerDetect............................................................................ 10-12 CornerEigenValsAndVecs.............................................................. 10-12 CornerMinEigenVal........................................................................ 10-13 FindCornerSubPix ......................................................................... 10-14 GoodFeaturesToTrack.................................................................... 10-16 Hough Transform Functions............................................................ 10-17 HoughLines.................................................................................... 10-17 HoughLinesSDiv ............................................................................ 10-18 Discussion................................................................................. 10-18 HoughLinesP ................................................................................. 10-19 Discussion................................................................................. 10-19 Image Statistics Functions.................................................................. 10-20 CountNonZero ............................................................................... 10-20 SumPixels ...................................................................................... 10-20

OpenCV Reference Manual

Contents

Mean .............................................................................................. Mean_StdDev ................................................................................ MinMaxLoc .................................................................................... Norm .............................................................................................. Moments ........................................................................................ GetSpatialMoment ......................................................................... GetCentralMoment......................................................................... GetNormalizedCentralMoment ...................................................... GetHuMoments.............................................................................. Pyramid Functions.............................................................................. PyrDown ........................................................................................ PyrUp ............................................................................................. PyrSegmentation ........................................................................... Morphology Functions ........................................................................ CreateStructuringElementEx ......................................................... ReleaseStructuringElement ........................................................... Erode ............................................................................................. Dilate.............................................................................................. MorphologyEx................................................................................ Distance Transform Function.............................................................. DistTransform................................................................................. Threshold Functions ........................................................................... AdaptiveThreshold ......................................................................... Threshold ....................................................................................... Flood Filling Function ......................................................................... FloodFill ......................................................................................... Histogram Functions........................................................................... CreateHist...................................................................................... ReleaseHist ................................................................................... MakeHistHeaderForArray .............................................................. QueryHistValue_1D ....................................................................... QueryHistValue_2D .......................................................................

10-21 10-21 10-22 10-22 10-24 10-25 10-25 10-26 10-27 10-28 10-28 10-28 10-29 10-30 10-30 10-31 10-31 10-32 10-33 10-34 10-34 10-36 10-36 10-38 10-40 10-40 10-41 10-41 10-42 10-42 10-43 10-43

OpenCV Reference Manual

Contents

QueryHistValue_3D ....................................................................... QueryHistValue_nD ....................................................................... GetHistValue_1D ........................................................................... GetHistValue_2D ........................................................................... GetHistValue_3D ........................................................................... GetHistValue_nD ........................................................................... GetMinMaxHistValue ..................................................................... NormalizeHist ................................................................................ ThreshHist ..................................................................................... CompareHist.................................................................................. CopyHist ........................................................................................ SetHistBinRanges.......................................................................... CalcHist ......................................................................................... CalcBackProject............................................................................. CalcBackProjectPatch.................................................................... CalcEMD........................................................................................ CalcContrastHist............................................................................ Pyramid Data Types ........................................................................... Histogram Data Types ........................................................................

10-44 10-44 10-45 10-45 10-46 10-46 10-47 10-47 10-48 10-48 10-49 10-50 10-50 10-51 10-52 10-54 10-55 10-56 10-57

OpenCV Reference Manual

Contents

PutText ........................................................................................... GetTextSize.................................................................................... Utility Reference ................................................................................. AbsDiff ........................................................................................... AbsDiffS ......................................................................................... MatchTemplate............................................................................... CvtPixToPlane................................................................................ CvtPlaneToPix................................................................................ ConvertScale ................................................................................. InitLineIterator ................................................................................ SampleLine.................................................................................... GetRectSubPix .............................................................................. bFastArctan.................................................................................... Sqrt ................................................................................................ bSqrt .............................................................................................. InvSqrt ........................................................................................... bInvSqrt ......................................................................................... bReciprocal.................................................................................... bCartToPolar .................................................................................. bFastExp........................................................................................ bFastLog ........................................................................................ RandInit ......................................................................................... bRand ............................................................................................ FillImage ........................................................................................ RandSetRange .............................................................................. KMeans..........................................................................................

14-75 14-75 14-76 14-76 14-77 14-77 14-80 14-80 14-81 14-82 14-83 14-84 14-84 14-85 14-85 14-86 14-86 14-87 14-87 14-88 14-88 14-89 14-89 14-90 14-90 14-91

Chapter 15 System Functions

1-3

OpenCV Reference Manual

Overview

Platforms Supported
The OpenCV software run on Windows platforms. The code and syntax used for function and variable declarations in this manual are written in the ANSI C style. However, versions of the OpenCV for different processors or operating systems may, of necessity, vary slightly.

About This Manual

This manual provides a background for the computer image processing concepts used in the OpenCV software. The manual includes two major parts, one is the Programmer Guide and the other is Reference. The fundamental concepts of each of the library components are extensively covered in the Programmer Guide. The Reference provides the user with specifications of each OpenCV function. The functions are combined into groups by their functionality (chapters 10 through 16). Each group of functions is described along with appropriate data types and macros, when applicable. The manual includes example codes of the library usage.

Manual Organization
This manual includes two principal parts: Programmer Guide and Reference. The Programmer Guide contains Overview (Chapter 1) that provides information on the OpenCV software, application area, overall functionality, the library relation to IPL, data types and error handling, along with manual organization and notational conventions. and the following functionality chapters: Chapter 2 Motion Analysis and Object Tracking comprising sections:

Background Subtraction. Describes basic functions that enable building statistical model of background for its further subtraction.

1-4

OpenCV Reference Manual

Overview

Motion Templates. Describes motion templates functions designed to generate motion template images that can be used to rapidly determine where a motion occurred, how it occurred, and in which direction it occurred. Cam Shift. Describes the functions implemented for realization of Continuously Adaptive Mean-SHIFT algorithm (CamShift) algorithm. Active Contours. Describes a function for working with active contours (snakes). Optical Flow. Describes functions used for calculation of optical flow implementing Lucas & Kanade, Horn & Schunck, and Block Matching techniques. Estimators. Describes a group of functions for estimating stochastic models state. Contour Retrieving. Describes contour retrieving functions. Features. Describes various fixed filters, primarily derivative operators (1st & 2nd Image Derivatives); feature detection functions; Hough Transform method of extracting geometric primitives from raster images. Image Statistics. Describes a set of functions that compute different information about images, considering their pixels as independent observations of a stochastic variable. Pyramids. Describes functions that support generation and reconstruction of Gaussian and Laplacian Pyramids. Morphology. Describes an expanded set of morphological operators that can be used for noise filtering, merging or splitting image regions, as well as for region boundary detection. Distance Transform. Describes the distance transform functions used for calculating the distance to an object.

Chapter 3

Image Analysis comprising sections:

1-5

OpenCV Reference Manual

Overview

Thresholding. Describes threshold functions used mainly for masking out some pixels that do not belong to a certain range, for example, to extract blobs of certain brightness or color from the image, and for converting grayscale image to bi-level or black-and-white image. Flood Filling. Describes the function that performs flood filling of a connected domain. Histogram. Describes functions that operate on multi-dimensional histograms. Contour Processing. Describes contour processing functions. Geometry. Describes functions from computational geometry field: line and ellipse fitting, convex hull, contour analysis. Eigen Objects. Describes functions that operate on eigen objects. Embedded HMM. Describes functions for using Embedded Hidden Markov Models (HMM) in face recognition task. Camera Calibration. Describes undistortion functions and camera calibration functions used for calculating intrinsic and extrinsic camera parameters. View Morphing. Describes functions for morphing views from two cameras. POSIT. Describes functions that together perform POSIT algorithm used to determine the six degree-of-freedom pose of a known tracked 3D rigid object. Gesture Recognition. Describes specific functions for the static gesture recognition technology.

Chapter 4

Structural Analysis comprising sections:

Chapter 5

Image Recognition comprising sections:

Chapter 6

3D Reconstruction comprising sections:

Chapter 7

Basic Structures and Operations comprising sections:

1-6

OpenCV Reference Manual

Overview

Chapter 8

Image Functions. Describes basic functions for manipulating raster images: creation, allocation, destruction of images. Fast pixel access macros are also described. Dynamic Data Structures. Describes several resizable data structures and basic functions that are designed to operate on these structures. Matrix Operations. Describes functions for matrix operations: basic matrix arithmetics, eigen problem solution, SVD, 3D geometry and recognition-specific functions. Drawing Primitives. Describes simple drawing functions intended mainly to mark out recognized or tracked features in Utility. Describes unclassified OpenCV functions.

Library Technical Organization and System Fuctions comprising sections:

Error Handling. Memory Management. Interaction With Low-Level Optimized Functions. User DLL Creation.

Reference contains the following chapters describing respective functions, data types and applicable macros: Chapter 9 Chapter 10 Chapter 11 Chapter 12 Chapter 13 Chapter 14 Chapter 15 Motion Analysis and Object Tracking Reference. Image Analysis Reference. Structural Analysis Reference. Image Recognition Reference. 3D Reconstruction Reference. Basic Structures and Operations Reference. System Functions Reference.

The manual also includes Appendix A that describes supported image attributes and operation modes, a Glossary of terms, a Bibliography, and an Index.

1-7

OpenCV Reference Manual

Overview

Function Descriptions
In Chapters 10 through 16, each function is introduced by name and a brief description of its purpose. This is followed by the function call sequence, definitions of its arguments, and more detailed explanation of the function purpose. The following sections are included in function description: Arguments Discussion Describes all the function arguments. Defines the function and describes the operation performed by the function. This section also includes descriptive equations.

Audience for This Manual

The manual is intended for all users of OpenCV: researchers, commercial software developers, government and camera vendors.

On-line Version
This manual is available in an electronic format (Portable Document Format, or PDF). To obtain a hard copy of the manual, print the file using the printing capability of Adobe* Acrobat*, the tool used for the on-line presentation of the document.

Related Publications
For more information about signal processing concepts and algorithms, refer to the books and materials listed in the Bibliography.

Notational Conventions
In this manual, notational conventions include:

Fonts used for distinction between the text and the code Naming conventions Function name conventions

1-8

OpenCV Reference Manual

Overview

Font Conventions
The following font conventions are used:
THIS TYPE STYLE

Used in the text for OpenCV constant identifiers; for example, CV_SEQ_KIND_GRAPH. Mixed with the uppercase in structure names as in CvContourTree; also used in function names, code examples and call statements; for example, int cvFindContours(). Variables in arguments discussion; for example, value, src.

This type style

Naming Conventions
The OpenCV software uses the following naming conventions for different items:

Constant identifiers are in uppercase; for example, CV_SEQ_KIND_GRAPH. All names of the functions used for image processing have the cv prefix. In code
examples, you can distinguish the OpenCV interface functions from the application functions by this prefix.

All OpenCV external functions names start with cv prefix, all structures names
start with Cv prefix.
NOTE. In this manual, the cv prefix in function names is always

used in the code examples. In the text, this prefix is usually omitted when referring to the function group. Prefix cvm is respectively omitted in Matrix Operations Functions. Each new part of a function name starts with an uppercase character, without underscore; for example, cvContourTree.

1-9

OpenCV Reference Manual

Overview

Function Name Conventions

The function names in the OpenCV library typically begin with cv prefix and have the following general format: cv <action> <target> <mod> () where
action

indicates the core functionality, for example, -Set-,

-Create-, -Convert-.

target

indicates the area where the image processing is being enacted,forexample, -Find Contours or -ApproxPoly. In a number of cases the target consists of two or more words, for example, -MatchContourTree. Some function names consist of an action or target only; for example, the functions cvUnDistort or cvAcc respectively.

mod

an optional field; indicates a modification to the core functionality of a function. For example, in the function name cvFindExtrinsicCameraParams_64d, _64d indicates that this particular function constant 64d values.

1-10

OpenCV Reference Manual

Overview

1-11

Motion Analysis and Object Tracking

Background Subtraction

This section describes basic functions that enable building statistical model of background for its further subtraction. In this chapter the term "background" stands for a set of motionless image pixels, that is, pixels that do not belong to any object, moving in front of the camera. This definition can vary if considered in other techniques of object extraction. For example, if a depth map of the scene is obtained, background can be determined as parts of scene that are located far enough from the camera. The simplest background model assumes that every background pixel brightness varies independently, according to normal distribution.The background characteristics can be calculated by accumulating several dozens of frames, as well as their squares. That means finding a sum of pixel values in the location S(x,y) and a sum of squares of the values Sq(x,y) for every pixel location. Then mean is calculated as collected, and standard deviation as
S ( x, y ) m ( x, y ) = --------------- , N

where N is the number of the frames

After that the pixel in a certain pixel location in certain frame is regarded as belonging to a moving object if condition abs ( m ( x, y ) p( x, y ) ) > C ( x, y ) is met, where C is a certain constant. If C is equal to 3, it is the well-known "three sigmas" rule. To obtain that background model, any objects should be put away from the camera for a few seconds, so that a whole image from the camera represents subsequent background observation. The above technique can be improved. First, it is reasonable to provide adaptation of background differencing model to changes of lighting conditions and background scenes, e.g., when the camera moves or some object is passing behind the front object.

2-1

Sq ( x, y ) S( x, y ) ( x, y ) = sqrt ------------------ --------------N N

OpenCV Reference Manual

Motion Analysis and Object Tracking

The simple accumulation in order to calculate mean brightness can be replaced with running average. Also, several techniques can be used to identify moving parts of the scene and exclude them in the course of background information accumulation. The techniques include change detection, e.g., via cvAbsDiff with cvThreshold, optical flow and, probably, others. The functions from the section (See Motion Analysis and Object Tracking Reference) are simply the basic functions for background information accumulation and they can not make up a complete background differencing module alone.

Motion Templates
The functions described in Motion Templates Functions section are designed to generate motion template images that can be used to rapidly determine where a motion occurred, how it occurred, and in which direction it occurred. The algorithms are based on papers by Davis and Bobick [Davis97] and Bradski and Davis [Bradsky00]. These functions operate on images that are the output of background subtraction or other image segmentation operations; thus the input and output image types are all grayscale, that is, have a single color channel.

Motion Representation and Normal Optical Flow Method

Motion Representation
Figure 2-1 (left) shows capturing a foreground silhouette of the moving object or person. Obtaining a clear silhouette is achieved through application of some of background subtraction techniques briefly described in the section on Background Subtraction. As the person or object moves, copying the most recent foreground silhouette as the highest values in the motion history image creates a layered history of the resulting motion; typically this highest value is just a floating point timestamp of time elapsing since the application was launched in milliseconds. Figure 2-1 (right)

2-2

OpenCV Reference Manual

Motion Analysis and Object Tracking

shows the result that is called the Motion History Image (MHI). A pixel level or a time delta threshold, as appropriate, is set such that pixel values in the MHI image that fall below that threshold are set to zero.
Figure 2-1 Motion History Image From Moving Silhouette

The most recent motion has the highest value, earlier motions have decreasing values subject to a threshold below which the value is set to zero. Different stages of creating and processing motion templates are described below.

A) Updating MHI Images

Generally, floating point images are used because system time differences, that is, time elapsing since the application was launched, are read in milliseconds to be further converted into a floating point number which is the value of the most recent silhouette. Then follows writing this current silhouette over the past silhouettes with subsequent thresholding away pixels that are too old (beyond a maximum mhiDuration) to create the MHI.

B) Making Motion Gradient Image

1. Start with the MHI image as shown in Figure 2-2(left). 2. Apply 3x3 Sobel operators X and Y to the image.

2-3

OpenCV Reference Manual

Motion Analysis and Object Tracking

3. If the resulting response at a pixel location (X,Y) is S x ( x, y ) to the Sobel operator X and S y ( x, y ) to the operator Y, then the orientation of the gradient is calculated as:
A ( x, y ) = arc tan S y ( ( x, y ) S x ( x, y ) ) ,

Figure 2-6

Block Diagram of CamShift Algorithm

Choose initial search window size and location

HSV Image

Set calculation region at search window center but larger in size than the search window

Color histogram lookup in calculation region

Color probability distribution image

Use (X,Y) to set search window center, 2*area1/2 to set size.

Find center of mass within the search window

Center search window at the center of mass and find area under it Report X, Y, Z, and Roll

YES

Converged?

CamShift operates on a 2D color probability distribution image produced from histogram back-projection (see the section on Histogram in Image Analysis). The core part of the CamShift algorithm is the Mean Shift algorithm. The Mean Shift part of the algorithm (gray area in Figure 2-6) is as follows: 1. Choose the search window size. 2. Choose the initial location of the search window.

2-10

OpenCV Reference Manual

Motion Analysis and Object Tracking

3. Compute the mean location in the search window. 4. Center the search window at the mean location computed in Step 3. 5. Repeat Steps 3 and 4 until the search window center converges, i.e., until it has moved for a distance less than the preset threshold.

Mass Center Calculation for 2D Probability Distribution

For discrete 2D image probability distributions, the mean location (the centroid) within the search window, that is computed at step 3 above, is found as follows: Find the zeroth moment

Find the first moment for x and y

x y

Mean search window location (the centroid) then is found as

M 10 M 01 x c = ------- ; y c = ------- , M 00 M 00

where I(x,y) is the pixel (probability) value in the position (x,y) in the image, and x and y range over the search window. Unlike the Mean Shift algorithm, which is designed for static distributions, CamShift is designed for dynamically changing distributions. These occur when objects in video sequences are being tracked and the object moves so that the size and location of the probability distribution changes in time. The CamShift algorithm adjusts the search window size in the course of its operation. Initial window size can be set at any reasonable value. For discrete distributions (digital data), the minimum window length or width is three. Instead of a set, or externally adapted window size, CamShift relies on the zeroth moment information, extracted as part of the internal workings of the algorithm, to continuously adapt its window size within or over each video frame.

M 10 =

x y

M 00 =

I ( x, y ) .

xI ( x, y ) ; M 01 =

yI ( x, y ) .

x y

2-11

OpenCV Reference Manual

Motion Analysis and Object Tracking

CamShift Algorithm
1. Set the calculation region of the probability distribution to the whole image. 2. Choose the initial location of the 2D mean shift search window. 3. Calculate the color probability distribution in the 2D region centered at the search window location in an ROI slightly larger than the mean shift window size. 4. Run Mean Shift algorithm to find the search window center. Store the zeroth moment (area or size) and center location. 5. For the next video frame, center the search window at the mean location stored in Step 4 and set the window size to a function of the zeroth moment found there. Go to Step 3. Figure 2-7 shows CamShift finding the face center on a 1D slice through a face and hand flesh hue distribution. Figure 2-8 shows the next frame when the face and hand flesh hue distribution has moved, and convergence is reached in two iterations.

2-12

OpenCV Reference Manual

Motion Analysis and Object Tracking

Figure 2-7

Cross Section of Flesh Hue Distribution

Step 1
250 200 150 100 50 1 0 4 250 200 150 100 50 1 0 7 10 13 16 19 22 3

Step 4

21 21 21

Step 2
250 200 150 100 50 1 0 4 250 200 150 100 50 1 0 7 10 13 16 19 22 3

Step 5

Step 3
250 200 150 100 50 1 0 3 250 200 150 100 50 1 0 5 7 9 11 13 15 17 19 21 23 3

Step 6

Rectangular CamShift window is shown behind the hue distribution, while triangle in front marks the window center. CamShift is shown iterating to convergence down the left then right columns.

2-13

OpenCV Reference Manual

Motion Analysis and Object Tracking

Figure 2-8

Flesh Hue Distribution (Next Frame)

Step 1
250 200 150 100 50 1 0 4 7 250 200 150 100 50 1 0 10 13 16 19 22 4 7

Step 2

Starting from the converged search location in Figure 2-7 bottom right, CamShift converges on new center of distribution in two iterations.

Calculation of 2D Orientation
The 2D orientation of the probability distribution is also easy to obtain by using the second moments in the course of CamShift operation, where the point (x,y) ranges over the search window, and I(x,y) is the pixel (probability) value at the point (x,y). Second moments are
x y 2 2

Then the object orientation, or direction of the major axis, is

M 11 2 ------- x c y c M 00 arc tan ---------------------------------------------------------M 20 M 02 2 2 ------- x c ------- y c M 00 M 00 = ------------------------------------------------------------------------------------ . 2

The first two eigenvalues, that is, length and width, of the probability distribution of the blob found by CamShift may be calculated in closed form as follows:

M 20 =

x I ( x, y ) , M 02 =

x I ( x, y ) .

x y

2-14

OpenCV Reference Manual

Motion Analysis and Object Tracking

Let

M20 M 11 2 a = ------- x c , b = 2 ------- xc y c M00 M 00

, and

M 02 2 c = ------- y c . M 00

Then length l and width w from the distribution centroid are

(a + c) + b + (a c) ------------------------------------------------------------- , 2 (a + c) b + (a c) ------------------------------------------------------------- . 2
2 2 2 2

l =

w =

When used in face tracking, the above equations give head roll, length, and width as marked in the source video image in Figure 2-9.
Figure 2-9 Orientation of Flesh Probability Distribution

Active Contours
This section describes a function for working with active contours, also called snakes. The snake was presented in [Kass88] as an energy-minimizing parametric closed curve guided by external forces. Energy function associated with the snake is E = E int + E ext , where E int is the internal energy formed by the snake configuration, E ext is the external energy formed by external forces affecting the snake. The aim of the snake is to find a location that minimizes energy.

2-15

OpenCV Reference Manual

Motion Analysis and Object Tracking

Let p 1, , p n be a discrete representation of a snake, that is, a sequence of points on an image plane. In OpenCV the internal energy function is the sum of the contour continuity energy and the contour curvature energy, as follows:
E int = E cont + E curv , E cont

where

is the contour continuity energy. This energy is E cont = d pi p i 1 , where d is the average distance between all pairs ( p i p i 1 ) . Minimizing E cont over all the snake points p 1, , p n , causes the snake points become more equidistant. is the contour curvature energy. The smoother the contour is, the less is the curvature energy. E curv = p i 1 2 p i + p i + 1 2 .
E ext = E img + E con ,

E curv

In [Kass88] external energy was represented as

E img

where

image energy and

E con -

energy of additional constraints.

Two variants of image energy are proposed: 1. 2.

E img = I , where I is the image intensity. In this case the snake is attracted to the bright lines of the image. E img = grad ( I )

. The snake is attracted to the image edges.

A variant of external constraint is described in [Kass88]. Imagine the snake points connected by springs with certain image points. Then the spring force k(x x0) produces the energy can be useful when
kx --------- . 2
2

This force pulls the snake points to fixed positions, which

snake points need to be fixed. OpenCV does not support this option now. Summary energy at every point can be written as
E i = i E cont, i + i E curv, i + i E img, i ,

(2.1)

where , , are the weights of every kind of energy. The full snake energy is the sum of E i over all the points. The meanings of
, ,

are as follows:

is responsible for contour continuity, that is, a big evenly spaced.

makes snake points more

2-16

OpenCV Reference Manual is responsible for snake corners, that is, a big between snake edges more obtuse.

Motion Analysis and Object Tracking

for a certain point makes the angle

is responsible for making the snake point more sensitive to the image energy, rather than to continuity or curvature. Only relative values of
, ,

in the snake point are relevant.

The following way of working with snakes is proposed:

create a snake with initial configuration; define weights

, ,

at every point;
, , ,

allow the snake to minimize its energy; evaluate the snake position. If required, adjust and repeat the previous step. and, possibly, image data,

There are three well-known algorithms for minimizing snake energy. In [Kass88] the minimization is based on variational calculus. In [Yuille89] dynamic programming is used. The greedy algorithm is proposed in [Williams92]. The latter algorithm is the most efficient and yields quite good results. The scheme of this algorithm for each snake point is as follows: 1. Use Equation (3.1) to compute E for every location from point neighborhood. Before computing E, each energy term E cont, E curv, E img must be normalized using formula E normalized = ( Eimg min ) ( max min ) , where max and min are maximal and minimal energy in scanned neighborhood. 2. Choose location with minimum energy. 3. Move snakes point to this location. 4. Repeat all the steps until convergence is reached. Criteria of convergence are as follows:

maximum number of iterations is achieved; number of points, moved at last iteration, is less than given threshold.
In [Williams92] the authors proposed a way, called high-level feedback, to adjust b coefficient for corner estimation during minimization process. Although this feature is not available in the implementation, the user may build it, if needed.

2-17

OpenCV Reference Manual

Motion Analysis and Object Tracking

Optical Flow
This section describes several functions for calculating optical flow between two images. Most papers devoted to motion estimation use the term optical flow. Optical flow is defined as an apparent motion of image brightness. Let I(x,y,t) be the image brightness that changes in time to provide an image sequence. Two main assumptions can be made: 1. Brightness I(x,y,t) smoothly depends on coordinates x, y in greater part of the image. 2. Brightness of every point of a moving or static object does not change in time. Let some object in the image, or some point of an object, move and after time dt the object displacement is (dx, dy). Using Taylor series for brightness I(x,y,t) gives the following:
I I I I ( x + dx, y + dy, t + dt ) = I ( x, y, t ) + ------ dx + ------ dy + ------ dt + , x y t

(2.2)

where are higher order terms. Then, according to Assumption 2:

I ( x + dx, y + dy, t + dt ) = I ( x, y, t ) ,

(2.3)

and
I I I ------ dx + ------ dy + ------ dt + = 0 . y t x

(2.4)

Dividing (18.3) by dt and defining

dx dy ------ = u , ------ = v dt dt

(2.5)

gives an equation
I I I ------ = ------ u + ------ v , t y x

(2.6)

usually called optical flow constraint equation, where u and v are components of optical flow field in x and y coordinates respectively. Since Equation (2.6) has more than one solution, more constraints are required. Some variants of further steps may be chosen. Below follows a brief overview of the options available.

2-18

OpenCV Reference Manual

Motion Analysis and Object Tracking

Lucas & Kanade Technique

Using the optical flow equation for a group of adjacent pixels and assuming that all of them have the same velocity, the optical flow computation task is reduced to solving a linear system. In a non-singular system for two pixels there exists a single solution of the system. However, combining equations for more than two pixels is more effective. In this case the approximate solution is found using the least square method. The equations are usually weighted. Here the following 2x2 linear system is used:
x, y x, y x, y

x, y

where W(x,y) is the Gaussian window. The Gaussian window may be represented as a composition of two separable kernels with binomial coefficients. Iterating through the system can yield even better results. It means that the retrieved offset is used to determine a new window in the second image from which the window in the first image is subtracted, while It is calculated.

Horn & Schunck Technique

Horn and Schunck propose a technique that assumes the smoothness of the estimated optical flow field [Horn81]. This constraint can be formulated as
image

This optical flow solution can deviate from the optical flow constraint. To express this deviation the following integral can be used:

The value S + C , where is a parameter, called Lagrangian multiplier, is to be minimized. Typically, a smaller must be taken for a noisy image and a larger one for a quite accurate image. To minimize S + C , a system of two second-order differential equations for the whole image must be solved:

C =

I 2 I I ----------------- u + ------ v + ------ dx dy . t x y image

S =

u -----x

u + -----y

v + -----x

v + -----y

W ( x, y )I x u +

W ( x, y )I x I y v =

W ( x, y )I x I y u +

W ( x, y )I y v =

W ( x , y )I y I t ,

W ( x , y )I x I t ,

( dx ) dy

(2.7)

(2.8)

2-19

OpenCV Reference Manual u u I I I I -------- + -------- = ------ u + ------ v + ------ ------ , 2 2 t x x y x y I I v v I I -------- + -------- = ------ u + ------ v + ------ ------ . 2 2 t x x y x y

Motion Analysis and Object Tracking

Iterative method could be applied for the purpose when a number of iterations are made for each pixel. This technique for two consecutive images seems to be computationally expensive because of iterations, but for a long sequence of images only an iteration for two images must be done, if the result of the previous iteration is chosen as initial approximation.

Block Matching
This technique does not use an optical flow equation directly. Consider an image divided into small blocks that can overlap. Then for every block in the first image the algorithm tries to find a block of the same size in the second image that is most similar to the block in the first image. The function searches in the neighborhood of some given point in the second image. So all the points in the block are assumed to move by the same offset that is found, just like in Lucas & Kanade method. Different metrics can be used to measure similarity or difference between blocks - cross correlation, squared difference, etc.

Estimators
This section describes group of functions for estimating stochastic models state. State estimation programs implement a model and an estimator. A model is analogous to a data structure representing relevant information about the visual scene. An estimator is analogous to the software engine that manipulates this data structure to compute beliefs about the world. The OpenCV routines provide two estimators: standard Kalman and condensation.

Models
Many computer vision applications involve repeated estimating, that is, tracking, of the system quantities that change over time. These dynamic quantities are called the system state. The system in question can be anything that happens to be of interest to a particular vision task.

(2.9)

2-20

OpenCV Reference Manual

Motion Analysis and Object Tracking

To estimate the state of a system, reasonably accurate knowledge of the system model and parameters may be assumed. Parameters are the quantities that describe the model configuration but change at a rate much slower than the state. Parameters are often assumed known and static. In OpenCV a state is represented with a vector. In addition to this output of the state estimation routines, another vector introduced is a vector of measurements that are input to the routines from the sensor data. To represent the model, two things are to be specified:

Estimated dynamics of the state change from one moment of time to the next Method of obtaining a measurement vector zt from the state.
Estimators
Most estimators have the same general form with repeated propagation and update phases that modify the state's uncertainty as illustrated in Figure 2-10.
Figure 2-10 Ongoing Discrete Kalman Filter Cycle

The time update projects the current state estimate ahead in time. The measurement update adjusts the projected estimate using an actual measurement at that time.

2-21

OpenCV Reference Manual

Motion Analysis and Object Tracking

An estimator should be preferably unbiased when the probability density of estimate errors has an expected value of 0. There exists an optimal propagation and update formulation that is the best, linear, unbiased estimator (BLUE) for any given model of the form. This formulation is known as the discrete Kalman estimator, whose standard form is implemented in OpenCV.

Kalman Filtering
The Kalman filter addresses the general problem of trying to estimate the state x of a discrete-time process that is governed by the linear stochastic difference equation
x k + 1 = Ax k + w k

(2.10)

with a measurement z, that is

z k = Hxk + vk

(2.11)

The random variables wk and vk respectively represent the process and measurement noise. They are assumed to be independent of each other, white, and with normal probability distributions
p ( w ) = N ( 0, Q ) , p ( w ) = N ( 0, R ) .

(2.12) (2.13)

The N x N matrix A in the difference equation (2.10) relates the state at time step k to the state at step k+1, in the absence of process noise. The M x N matrix H in the measurement equation (2.11) relates the state to the measurement zk.
If Xk denotes a priori state estimate at step k provided the process prior to step k is known, and Xk denotes a posteriori state estimate at step k provided measurement zk is known, then a priori and a posteriori estimate errors can be defined as
ek = x k X k ek = x k X k

. The a priori estimate error covariance is then

Pk =
T E [ ekek ]

P k = E [ e k ek ]

and the a

posteriori estimate error covariance is

The Kalman filter estimates the process by using a form of feedback control: the filter estimates the process state at some time and then obtains feedback in the form of noisy measurements. As such, the equations for the Kalman filter fall into two groups: time update equations and measurement update equations. The time update equations are responsible for projecting forward in time the current state and error covariance

2-22

OpenCV Reference Manual

Motion Analysis and Object Tracking

estimates to obtain the a priori estimates for the next time step. The measurement update equations are responsible for the feedback, that is, for incorporating a new measurement into the a priori estimate to obtain an improved a posteriori estimate. The time update equations can also be viewed as predictor equations, while the measurement update equations can be thought of as corrector equations. Indeed, the final estimation algorithm resembles that of a predictor-corrector algorithm for solving numerical problems as shown in Figure 2-10. The specific equations for the time and measurement updates are presented below. Time Update Equations
Xk + 1 = Ak Xk , P k + 1 = A k P k A k + Qk .
T

Measurement Update Equations:

K k = P k H k ( H k Pk H k + R k )
T T 1

Xk = Xk + Kk( zk Hk Xk) , P k = ( I K k H k )P k ,

where K is the so-called Kalman gain matrix and I is the identity operator. See CvKalman in Motion Analysis and Object Tracking Reference.

ConDensation Algorithm
This section describes the ConDensation (conditional density propagation) algorithm, based on factored sampling. The main idea of the algorithm is using the set of randomly generated samples for probability density approximation. For simplicity, general principles of ConDensation algorithm are described below for linear stochastic dynamical system:
x k + 1 = Ax k + w k

(2.14)

with a measurement Z. To start the algorithm, a set of samples Xn must be generated. The samples are randomly generated vectors of states. The function ConDensInitSampleSet does it in OpenCV implementation.

2-23

OpenCV Reference Manual

Motion Analysis and Object Tracking

During the first phase of the condensation algorithm every sample in the set is updated according to Equation (3.14). Further, when the vector of measurement Z is obtained, the algorithm estimates conditional probability densities of every sample P ( Xn Z ) . The OpenCV implementation of the ConDensation algorithm enables the user to define various probability density functions. There is no such special function in the library. After the probabilities are calculated, the user may evaluate, for example, moments of tracked process at the current time step. If dynamics or measurement of the stochastic system is non-linear, the user may update the dynamics (A) or measurement (H) matrices, using their Taylor series at each time step. See CvConDensation in Motion Analysis and Object Tracking Reference.

2-24

Image Analysis
Contour Retrieving
This section describes contour retrieving functions. Below follow descriptions of:

several basic functions that retrieve contours from the binary image and store them
in the chain format;

functions for polygonal approximation of the chains.

Basic Definitions
Most of the existing vectoring algorithms, that is, algorithms that find contours on the raster images, deal with binary images. A binary image contains only 0-pixels, that is, pixels with the value 0, and 1-pixels, that is, pixels with the value 1. The set of connected 0- or 1-pixels makes the 0-(1-) component. There are two common sorts of connectivity, the 4-connectivity and 8-connectivity. Two pixels with coordinates (x, y) and ( x, y) are called 4-connected if, and only if, x x + y y = 1 and 8-connected if, and only if, max ( x x , y y ) = 1 . Figure 3-1 shows these relations.:
Figure 3-1 Pixels Connectivity Patterns

Pixels, 8-connected to the black one

Pixels, 4- and 8-connected to the black one

3-1

OpenCV Reference Manual

Image Analysis

Using this relationship, the image is broken into several non-overlapped 1-(0-) 4-connected (8-connected) components. Each set consists of pixels with equal values, that is, all pixels are either equal to 1 or 0, and any pair of pixels from the set can be linked by a sequence of 4- or 8-connected pixels. In other words, a 4-(8-) path exists between any two points of the set. The components shown in Figure 3-2 may have interrelations.
Figure 3-2 Hierarchical Connected Components

1-components W1, W2, and W3 are inside the frame (0-component B1), that is, directly surrounded by B1. 0-components B2 and B3 are inside W1. 1-components W5 and W6 are inside B4, that is inside W3, so these 1-components are inside W3 indirectly. However, neither W5 nor W6 enclose one another, which means they are on the same level. In order to avoid a topological contradiction, 0-pixels must be regarded as 8-(4-) connected pixels in case 1-pixels are dealt with as 4-(8-) connected. Throughout this document 8-connectivity is assumed to be used with 1-pixels and 4-connectivity with 0-pixels.

3-2

OpenCV Reference Manual

Image Analysis

Since 0-components are complementary to 1-components, and separate 1-components are either nested to each other or their internals do not intersect, the library considers 1-components only and only their topological structure is studied, 0-pixels making up the background. A 0-component directly surrounded by a 1-component is called the hole of the 1-component. The border point of a 1-component could be any pixel that belongs to the component and has a 4-connected 0-pixel. A connected set of border points is called the border. Each 1-component has a single outer border that separates it from the surrounding 0-component and zero or more hole borders that separate the 1-component from the 0-components it surrounds. It is obvious that the outer border and hole borders give a full description of the component. Therefore all the borders, also referred to as contours, of all components stored with information about the hierarchy make up a compressed representation of the source binary image. See Reference for description of the functions FindContours, StartFindContours, and FindNextContour that build such a contour representation of binary images.

Contour Representation
The library uses two methods to represent contours. The first method is called the Freeman method or the chain code. For any pixel all its neighbors with numbers from 0 to 7 can be enumerated:
Figure 3-3 Contour Representation in Freeman Method

3 2 1 4 0 5 6 7
The 0-neighbor denotes the pixel on the right side, etc. As a sequence of 8-connected points, the border can be stored as the coordinates of the initial point, followed by codes (from 0 to 7) that specify the location of the next point relative to the current one (see Figure 3-4).

3-3

OpenCV Reference Manual

Image Analysis

Figure 3-4

Freeman Coding of Connected Components

Initial Point Chain Code for the Curve: 34445670007654443

The chain code is a compact representation of digital curves and an output format of the contour retrieving algorithms described below. Polygonal representation is a different option in which the curve is coded as a sequence of points, vertices of a polyline. This alternative is often a better choice for manipulating and analyzing contours over the chain codes; however, this representation is rather hard to get directly without much redundancy. Instead, algorithms that approximate the chain codes with polylines could be used.

Contour Retrieving Algorithm

Four variations of algorithms described in [Suzuki85] are used in the library to retrieve borders. 1. The first algorithm finds only the extreme outer contours in the image and returns them linked to the list. Figure 3-2 shows these external boundaries of W1, W2, and W3 domains. 2. The second algorithm returns all contours linked to the list. Figure 3-2 shows the total of 8 such contours. 3. The third algorithm finds all connected components by building a two-level hierarchical structure: on the top are the external boundaries of 1-domains and every external boundary contains a link to the list of holes of the corresponding component. The third algorithm returns all the connected components as a two-level hierarchical structure: on the top are the external boundaries of 1-domains and every external boundary contour header contains

3-4

OpenCV Reference Manual

Image Analysis

a link to the list of holes in the corresponding component. The list can be accessed via v_next field of the external contour header. Figure 3-2 shows that W2, W5, and W6 domains have no holes; consequently, their boundary contour headers refer to empty lists of hole contours. W1 domain has two holes - the external boundary contour of W1 refers to a list of two hole contours. Finally, W3 external boundary contour refers to a list of the single hole contour. 4. The fourth algorithm returns the complete hierarchical tree where all the contours contain a list of contours surrounded by the contour directly, that is, the hole contour of W3 domain has two children: external boundary contours of W5 and W6 domains. All algorithms make a single pass through the image; there are, however, rare instances when some contours need to be scanned more than once. The algorithms do line-by-line scanning. Whenever an algorithm finds a point that belongs to a new border the border following procedure is applied to retrieve and store the border in the chain format. During the border following procedure the algorithms mark the visited pixels with special positive or negative values. If the right neighbor of the considered border point is a 0-pixel and, at the same time, the 0-pixel is located in the right hand part of the border, the border point is marked with a negative value. Otherwise, the point is marked with the same magnitude but of positive value, if the point has not been visited yet. This can be easily determined since the border can cross itself or tangent other borders. The first and second algorithms mark all the contours with the same value and the third and fourth algorithms try to use a unique ID for each contour, which can be used to detect the parent of any newly met border.

Features
Fixed Filters
This section describes various fixed filters, primarily derivative operators.

3-5

OpenCV Reference Manual

Image Analysis

Sobel Derivatives
Figure 3-5 shows first x derivative Sobel operator. The grayed bottom left number indicates the origin in a p-q coordinate system. The operator can be expressed as a polynomial and decomposed into convolution primitives.

Figure 3-5

First x Derivative Sobel Operator

2 q 1 0

1 2 1 0

0 0 0 1 p

-1 -2 -1 2 1 1 (1+q) 1

1 (1+q)

-1

(1+p)

(1-p)

For example, first x derivative Sobel operator may be expressed as a polynomial 2 2 2 2 2 2 2 1 + 2 q + q p 2 p q p q = ( 1 + q ) ( 1 p ) = ( 1 + q ) ( 1 + q ) ( 1 + p ) ( 1 p ) and decomposed into convolution primitives as shown in Figure 3-5. This may be used to express a hierarchy of first x and y derivative Sobel operators as follows:

-----x -----x

(1 + p)

n1 n

(1 + q) (1 p)
n1

(3.1) (3.2)

(1 + p) (1 + q )

(1 q)

for

n>0.

Figure 3-6 shows the Sobel first derivative filters of equations (3.1) and (3.2) for n = 2, 4. The Sobel filter may be decomposed into simple add-subtract convolution primitives.

3-6

OpenCV Reference Manual

Image Analysis

Figure 3-6

First Derivative Sobel Operators for n=2 and n= 4

Filter n=2 1 2 1
dx

Differentiate 1 1

Average

0 0 0
dy

*
0 1 1 1 * 1

* 1

0 1

0 2

Second derivative Sobel operators can be expressed in polynomial decomposition similar to equations (3.1) and (3.2). The second derivative equations are:

-------2 x -------2 y
2 2

(1 + p)

(1 + q) (1 p)

,
2

(3.3)

(1 + p)

(1 + q)

(1 q)

(3.4)

n1 n1 (1 + p) (1 + q ) (1 p)(1 q) ------------xy for n = 2, 3,.

(3.5)

3-7

OpenCV Reference Manual

Image Analysis

Figure 3-7 shows the filters that result for n = 2 and 4. Just as shown in Figure 3-6, these filters can be decomposed into simple add-subtract separable convolution operators as indicated by their polynomial form in the equations.
Figure 3-7 Sobel Operator Second Order Derivators for n = 2 and n = 4

The polynomial decomposition is shown above each operator.

2/x2 = (1+q)2(1-p)2 2/y2 = (1+p)2(1-q)2 2/xy = (1+q)(1+p)(1-q)(1-p)

1 2 1

-2 -4 -2

1 2 1

1 -2 1

2 -4 2

1 -2 1

-1 0 1

0 0 0

1 0 -1

2/x2 = (1+p)2(1+q)4(1-p)2

2/y2 = (1+q)2(1+p)4(1-q)2

1 4 6 4 1

0 0 0 0 0

-2 -4 -12 -8 -2

0 0 0 0 0

1 4 6 4 1

1 0 -2 0 1

4 0 -8 0 4

6 0 -12 0 6

4 0 -8 0 4

1 0 -2 0 1

2/xy = (1+p)3(1+q)3(1-p)(1-q)

-1 -2 0 2 1

-2 -4 0 4 2

0 0 0 0 0

2 4 0 -4 -2

1 2 0 -2 -1

3-8

OpenCV Reference Manual

Image Analysis

Third derivative Sobel operators can also be expressed in the polynomial decomposition form:

-------3 x -------3 y
3 3

(1 + p)

(1 + q) (1 p)

(3.6)

(1 + p) (1 + q)

(1 q)

,
n1

(3.7)
(1 q) ,

for n =3, 4,. The third derivative filter needs to be applied only for the cases n = 4 and general.

Optimal Filter Kernels with Floating Point Coefficients

First Derivatives
Table 3-1 gives coefficients for five increasingly accurate x derivative filters, the y filter derivative coefficients are just column vector versions of the x derivative filters.
Table 3-1 Coefficients for Accurate First Derivative Filters
Anchor 0 0 0 0 0 0.74038 0.833812 0.88464 0.914685 0.934465 DX Mask Coefficients -0.12019 -0.229945 -0.298974 -0.346228 -0.378736 0.0420264 0.0949175 0.138704 0.173894 -0.0178608 -0.0453905 -0.0727275 0.0086445 0.0239629 -0.00459622

Five increasingly accurate separable x derivative filter coefficients. The table gives half coefficients only. The full table can be obtained by mirroring across the central anchor coefficient. The greater the number of coefficients used, the less distortion from the ideal derivative filter.

---------------2 x y

( 1 p) ( 1 + p )

(1 + q )

(3.8)

( 1 p) ( 1 + p )

(1 + q )

(1 q)

(3.9)

3-9

OpenCV Reference Manual

Image Analysis

Second Derivatives
Table 3-2 gives coefficients for five increasingly accurate x second derivative filters. The y second derivative filter coefficients are just column vector versions of the x second derivative filters.
Table 3-2 Coefficients for Accurate Second Derivative Filters
Anchor -2.20914 -2.71081 -2.92373 -3.03578 -3.10308 DX Mask Coefficients 1.10457 1.48229 1.65895 1.75838 1.81996 -0.126882 -0.224751 -0.291985 -0.338852 0.0276655 0.0597665 0.088077 -0.00827 -0.0206659 0.00301915

The table gives half coefficients only. The full table can be obtained by mirroring across the central anchor coefficient. The greater the number of coefficients used, the less distortion from the ideal derivative filter.

Laplacian Approximation
The Laplacian operator is defined as the sum of the second derivatives x and y:
L = -------- + -------- . 2 2 x y
2 2

(3.10)

Thus, any of the equations defined in the sections for second derivatives may be used to calculate the Laplacian for an image.

Feature Detection
A set of Sobel derivative filters may be used to find edges, ridges, and blobs, especially in a scale-space, or image pyramid, situation. Below follows a description of methods in which the filter set could be applied.

is the first derivative in the direction x just as Dy. is the second derivative in the direction x just as Dyy. is the partial derivative with respect to x and y. is the third derivative in the direction x just as Dyyy.

Dxx Dxy

Dxxx

3-10

OpenCV Reference Manual

Image Analysis

Dxxy

and Dxyy are the third partials in the directions x, y.

Corner Detection
Method 1 Corners may be defined as areas where level curves multiplied by the gradient magnitude raised to the power of 3 assume a local maximum
D x D yy + D y D xx 2 D x D y D xy .
2 2

(3.11)

Method 2 Sobel first derivative operators are used to take the derivatives x and y of an image, after which a small region of interest is defined to detect corners in. A 2x2 matrix of the sums of the derivatives x and y is subsequently created as follows:

C =

The eigenvalues are found by solving det ( C I ) = 0 , where is a column vector of the eigenvalues and I is the identity matrix. For the 2x2 matrix of the equation above, the solutions may be written in a closed form:
D x + D y ( D x + D y ) 4 ( D x Dy ( D x D y ) = ------------------------------------------------------------------------------------------------------------------------------------------------------------------- . 2

If 1 , 2 > t , where t is some threshold, then a corner is found at that location. This can be very useful for object or shape recognition.

Canny Edge Detector

Edges are the boundaries separating regions with different brightness or color. J.Canny suggested in [Canny86] an efficient method for detecting edges. It takes grayscale image on input and returns bi-level image where non-zero pixels mark detected edges. Below the 4-stage algorithm is described.

3-11

DxDy
2 2

D x Dy Dy
2

(3.12)

2 2

Image Analysis

The Gaussian image pyramid on the left is used to create the Laplacian pyramid in the center, which is used to reconstruct the Gaussian pyramid and the original image on the right. In the figure, I is the original image, G is the Gaussian image, L is the Laplacian image. Subscripts denote level of the pyramid. A Gaussian kernel g is used to convolve the image before down-sampling or after up-sampling. Image Segmentation by Pyramid Computer vision uses pyramid based image processing techniques on a wide scale now. The pyramid provides a hierarchical smoothing, segmentation, and hierarchical computing structure that supports fast analysis and search algorithms. P. J. Burt suggested a pyramid-linking algorithm as an effective implementation of a combined segmentation and feature computation algorithm [Burt81]. This algorithm, described also in [Jahne97], finds connected components without preliminary threshold, that is, it works on grayscale image. It is an iterative algorithm. Burts algorithm includes the following steps: 1. Computation of the Gaussian pyramid. 2. Segmentation by pyramid-linking. 3. Averaging of linked pixels. Steps 2 and 3 are repeated iteratively until a stable segmentation result is reached. After computation of the Gaussian pyramid a son-father relationship is defined between nodes (pixels) in adjacent levels. The following attributes may be defined for every node (i,j) on the level l of the pyramid:
c[i,j,l][t] a[i,j,l][t]

is the value of the local image property, e.g., intensity; is the area over which the property has been computed; pointer to the nodes father, which is at level l+1;

p[[i,j,l][t] is s[i,j,l][t]

is the segment property, the average value for the entire segment containing the node. The letter t stands for the iteration number
(t 0) .

For

t = 0 , c [ i, j, l ] [ 0 ] = G i, j .

For every node (i,j) at level l there are 16 candidate son nodes at level l-1 (i,j), where

3-17

OpenCV Reference Manual i' { 2 i 1, 2 i, 2 i + 1, 2 i + 2 } , j' { 2 j 1, 2 j, 2 j + 1, 2 j + 2 } .

Image Analysis

(3.14)

For every node (i,j) at level l there are 4 candidate father nodes at level l+1 (i,j), (see Figure 3-10), where
i'' { ( i 1 ) 2, i + 1 ) 2 } , j'' { ( j 1 ) 2, j + 1 ) 2 } .

(3.15)

Son-father links are established for all nodes below the top of pyramid for every iteration t. Let d[n][t] be the absolute difference between the c value of the node (i,j)at level l and its nth candidate father, then
p [ i, j, l ] [ t ] = arg min d [ n ] [ t ]
1n4

(3.16)

Figure 3-10 Connections between Adjacent Pyramid Levels

[i" , j" , l + 1]

[i , j , l ]

After the son-father relationship is defined, the t, c, and a values are computed from bottom to the top for the 0 l n as where sum is calculated over all (i,j)node sons, as indicated by the links p in (3.16).

3-18

a [ i, j, 0 ] [ t ] = 1 , c [ i, j, 0 ] [ t ] = c [ i, j, 0 ] [ 0 ] , a [ i, j, l ] [ t ] =

a [ i ', j', l 1 ] [ t ] ,

OpenCV Reference Manual

Image Analysis

If a [ i, j, l ] [ t ] > 0 then c [ i, j, l ] [ t ] = ( [ i', j', l' 1 ] [ t ] c [ i', j', l 1 ] [ t ] ) a [ i, j, l ] [ t ] , but if a [ i, j, 0] [ t ] = 0 , the node has no sons, c [ i, j, 0] [ t ] is set to the value of one of its candidate sons selected at random. No segment values are calculated in the top down order. The value of the initial level L is an input parameter of the algorithm. At the level L the segment value of each node is set equal to its local property value:
s [ i, j, L ] [ t ] = c [ i, j, L ] [ t ] . l<L

For lower levels

s [ i, j, l ] [ t ] =

each node value is just that of its father c [ i'', j'', l + 1 ] [ t ] .

Here node (i,j) is the father of (i,j), as established in Equation (3.16). After this the current iteration t finishes and the next iteration t + 1 begins. Any changes in pointers in the next iteration result in changes in the values of local image properties. The iterative process is continued until no changes occur between two successive iterations. The choice of L only determines the maximum possible number of segments. If the number of segments less than the numbers of nodes at the level L, the values of c [ i, j, L ] [ t ] are clustered into a number of groups equal to the desired number of segments. The group average value is computed from the c values of its members, weighted by their areas a, and replaces the value c for each node in the group. See Pyramid Data Types in Image Analysis Reference.

Morphology
This section describes an expanded set of morphological operators that can be used for noise filtering, merging or splitting image regions, as well as for region boundary detection. Mathematical Morphology is a set-theory method of image analysis first developed by Matheron and Serra at the Ecole des Mines, Paris [Serra82]. The two basic morphological operations are erosion, or thinning, and dilation, or thickening. All operations involve an image A, called the object of interest, and a kernel element B, called the structuring element. The image and structuring element could be in any number of dimensions, but the most common use is with a 2D binary image, or with a

3-19

OpenCV Reference Manual

Image Analysis

3D grayscale image. The element B is most often a square or a circle, but could be any shape. Just like in convolution, B is a kernel or template with an anchor point. Figure 3-11 shows dilation and erosion of object A by B. The element B is rectangular with an anchor point at upper left shown as a dark square.

Figure 3-11 Dilation and Erosion of A by B

Dilation by B Erosion by B

If B t is the translation of B around the image, then dilation of object A by structuring element B is

AB =

t : Bt A 0

It means every pixel is in the set, if the intersection is not null. That is, a pixel under the anchor point of B is marked on, if at least one pixel of B is inside of A.
A nB

indicates the dilation is done n times.

Erosion of object A by structuring element B is

A B = {t :Bt A } .

That is, a pixel under the anchor of B is marked on, if B is entirely within A.

3-20

OpenCV Reference Manual

Image Analysis

A nB indicates the erosion is done n times and can be useful in finding boundary of A:

A ,

the

A = A ( A nB ) .

Opening of A by B is
A B = ( A nB ) nB .

(3.17)

Closing of A by B is
A B = ( A nB ) nB ,

(3.18)

where n > 0.

Flat Structuring Elements for Gray Scale

Erosion and dilation can be done in 3D, that is, with gray levels. 3D structuring elements can be used, but the simplest and the best way is to use a flat structuring element B as shown in Figure 3-12. In the figure, B has an anchor slightly to the right of the center as shown by the dark mark on B. Figure 3-12 shows 1D cross-section of both dilation and erosion of a gray level image A by a flat structuring element B.

3-21

OpenCV Reference Manual

Thresholding
This section describes threshold functions group. Thresholding functions are used mainly for two purposes: masking out some pixels that do not belong to a certain range, for example, to extract blobs of certain brightness or color from the image; converting grayscale image to bi-level or black-and-white image. Usually, the resultant image is used as a mask or as a source for extracting higher-level topological information, e.g., contours (see Active Contours), skeletons (see Distance Transform), lines (see Hough Transform functions), etc. Generally, threshold is a determined function t(x,y) on the image:
t ( x, y ) = A ( p ( x, y ) ) , f ( x, y, p ( x, y ) ) = true B ( p ( x, y ) ) , f ( x, y, p ( x, y ) ) = false

The predicate function f(x,y,p(x,y)) is typically represented as g(x,y) < p(x,y) < h(x,y), where g and h are some functions of pixel value and in most cases they are simply constants. There are two basic types of thresholding operations. The first type uses a predicate function, independent from location, that is, g(x,y) and h(x,y)are constants over the image. However, for concrete image some optimal, in a sense, values for the constants can be calculated using image histograms (see Histogram) or other statistical criteria (see Image Statistics). The second type of the functions chooses g(x,y) and h(x,y)depending on the pixel neigborhood in order to extract regions of varying brightness and contrast. The functions, described in this chapter, implement both these approaches. They support single-channel images with depth IPL_DEPTH_8U, IPL_DEPTH_8S or IPL_DEPTH_32F and can work in-place.

3-24

OpenCV Reference Manual

Image Analysis

Flood Filling
This section describes the function performing flood filling of a connected domain. Flood filling means that a group of connected pixels with close values is filled with, or is set to, a certain value. The flood filling process starts with some point, called seed, that is specified by function caller and then it propagates until it reaches the image ROI boundary or cannot find any new pixels to fill due to a large difference in pixel values. For every pixel that is just filled the function analyses:

4 neighbors, that is, excluding the diagonal neighbors; this kind of connectivity is
called 4-connectivity, or

8 neighbors, that is, including the diagonal neighbors; this kind of connectivity is
called 8-connectivity. The parameter connectivity of the function specifies the type of connectivity. The function can be used for:

segmenting a grayscale image into a set of uni-color areas, marking each connected component with individual color for bi-level images.
The function supports single-channel images with the depth IPL_DEPTH_8U or
IPL_DEPTH_32F.

Histogram
This section describes functions that operate on multi-dimensional histograms. Histogram is a discrete approximation of stochastic variable probability distribution. The variable can be either a scalar or a vector. Histograms are widely used in image processing and computer vision. For example, one-dimensional histograms can be used for:

grayscale image enhancement determining optimal threshold levels (see Thresholding) selecting color objects via hue histograms back projection (see CamShift), and
other operations. Two-dimensional histograms can be used for:

3-25

OpenCV Reference Manual

Image Analysis

analyzing and segmenting color images, normalized to brightness (e.g. red-green

or hue-saturation images),

analyzing and segmenting motion fields (x-y or magnitude-angle histograms), analyzing shapes (see CalcPGH in Geometry Functions section of Structural
Analysis Reference) or textures. Multi-dimensional histograms can be used for:

content based retrieval (see the function CalcPGH), bayesian-based object recognition (see [Schiele00]).
To store all the types of histograms (1D, 2D, nD), OpenCV introduces special structure CvHistogram described in Example 10-2 in Image Analysis Reference. Any histogram can be stored either in a dense form, as a multi-dimensional array, or in a sparse form with a balanced tree used now. However, it is reasonable to store 1D or 2D histograms in a dense form and 3D and higher dimensional histograms in a sparse form. The type of histogram representation is passed into histogram creation function and then it is stored in type field of CvHistogram. The function MakeHistHeaderForArray can be used to process histograms allocated by the user with Histogram Functions.

Histograms and Signatures

Histograms represent a simple statistical description of an object, e.g., an image. The object characteristics are measured during iterating through that object: for example, color histograms for an image are built from pixel values in one of the color spaces. All possible values of that multi-dimensional characteristic are further quantized on each coordinate. If the quantized characteristic can take different k1 values on the first coordinate, k2 values on second, and kn on the last one, the resulting histogram has the size
size =
n

ki .
i=1

3-26

OpenCV Reference Manual

Image Analysis

The histogram can be viewed as a multi-dimensional array. Each dimension corresponds to a certain object feature. An array element with coordinates [i1, i2 in], otherwise called a histogram bin, contains a number of measurements done for the object with quantized value equal to i1 on first coordinate, i2 on the second coordinate, and so on. Histograms can be used to compare respective objects:
1

D ( H, K ) =

(h k) A(h k) .

But these methods suffer from several disadvantages. The measure D L 1 sometimes gives too small difference when there is no exact correspondence between histogram bins, that is, if the bins of one histogram are slightly shifted. On the other hand, D L gives too large difference due to cumulative property. 2 Another drawback of pure histograms is large space required, especially for higher-dimensional characteristics. The solution is to store only non-zero histogram bins or a few bins with the highest score. Generalization of histograms is termed signature and defined in the following way: 1. Characteristic values with rather fine quantization are gathered. 2. Only non-zero bins are dynamically stored. This can be implemented using hash-tables, balanced trees, or other sparse structures. After processing, a set of clusters is obtained. Each of them is characterized by the coordinates and weight, that is, a number of measurements in the neighborhood. Removing clusters with small weight can further reduce the signature size. Although these structures cannot be compared using formulas written above, there exists a robust comparison method described in [RubnerJan98] called Earth Mover Distance. Earth Mover Distance (EMD) Physically, two signatures can be viewed as two systems - earth masses, spread into several localized pieces. Each piece, or cluster, has some coordinates in space and weight, that is, the earth mass it contains. The distance between two systems can be measured then as a minimal work needed to get the second configuration from the first or vice versa. To get metric, invariant to scale, the result is to be divided by the total mass of the system.

D L ( H, K ) =

hi ki

, or

3-27

OpenCV Reference Manual

Image Analysis

Mathematically, it can be formulated as follows. Consider m suppliers and n consumers. Let the capacity of ith supplier be xi and the capacity of jth consumer be yj. Also, let the ground distance between ith supplier and jth consumer be cij. The following restrictions must be met:
x i 0, y j 0, c i, j 0 , xi yj ,

Then the task is to find the flow matrix f ij , where f ij is the amount of earth, transferred from ith supplier to jth consumer. This flow must satisfy the restrictions below:
f i, j 0 , f i, j x i ,

and minimize the overall cost:

EMD ( x, y ) = ------------------------------------- . f i, j
i j i j

The task of finding the optimal flow is a well known transportation problem, which can be solved, for example, using the simplex method.

i j

0 i < m, 0 j < n .

f ij

f i, j = y

min

ci, j, fi, j .

i j

is the optimal flow, then Earth Mover Distance is defined as

c i, j f i, j

3-28

OpenCV Reference Manual

Image Analysis

Example Ground Distances

As shown in the section above, physically intuitive distance between two systems can be found if the distance between their elements can be measured. The latter distance is called ground distance and, if it is a true metric, then the resultant distance between systems is a metric too. The choice of the ground distance depends on the concrete task as well as the choice of the coordinate system for the measured characteristic. In [RubnerSept98], [RubnerOct98] three different distances are considered. 1. The first is used for human-like color discrimination between pictures. CIE Lab model represents colors in a way when a simple Euclidean distance gives true human-like discrimination between colors. So, converting image pixels into CIE Lab format, that is, representing colors as 3D vectors (L,a,b), and quantizing them (in 25 segments on each coordinate in [RubnerSept98]), produces a color-based signature of the image. Although in experiment, made in [RubnerSept98], the maximal number of non-zero bins could be 25x25x25 = 15625, the average number of clusters was ~8.8, that is, resulting signatures were very compact. 2. The second example is more complex. Not only the color values are considered, but also the coordinates of the corresponding pixels, which makes it possible to differentiate between pictures of similar color palette but representing different color regions placements: e.g., green grass at the bottom and blue sky on top vs. green forest on top and blue lake at the bottom. 5D 12 space is used and metric is: [ ( L ) 2 + ( a ) 2 + ( b ) 2 + ( ( x )2 + ( y ) 2 ) ] , where regulates importance of the spatial correspondence. When = 0, the first metric is obtained. 3. The third example is related to texture metrics. In the example Gabor transform is used to get the 2D vector texture descriptor (l,m), which is a log-polar characteristic of the texture. Then, no-invariance ground distance is defined as: d ( ( l 1, m1 ), ( l 2, m 2 ) ) = l + m , l = min ( l 1 l 2 , L l1 l 2 ) , m = m 1 m 2 , where is the scale parameter of Gabor transform, L is the number of different angles used (angle resolution), and M is the number of scales used (scale resolution). To get invariance to scale and rotation, the user may calculate minimal EMD for several scales and rotations:
( l 1, m 1 ), ( l 2, m 2 ) ,

3-29

OpenCV Reference Manual EMD ( t 1, t 2 ) = min EMD ( t 1, t 2, l0, m 0 ) ,

0 l0 < L M < m 0 < M

Image Analysis

where d is measured as in the previous case, but

and

look slightly different: .

l = min ( l 1 l 2 + l 0 ( mod L ) , L l 1 l 2 + l 0 ( mod L ) ) , m = m 1 m 2 + m 0

Lower Boundary for EMD

If ground distance is metric and distance between points can be calculated via the norm of their difference, and total suppliers capacity is equal to total consumers capacity, then it is easy to calculate lower boundary of EMD because:
i j i j

i j

As it can be seen, the latter expression is the distance between the mass centers of the systems. Poor candidates can be efficiently rejected using this lower boundary for EMD distance, when searching in the large image database.

x p

y q

3-30

j i

p i q i f i, j =

f i, j p i

f i, j qj

i j

c i, j, f i, j =

p i qj f i, j =

p i q j f i, j

Structural Analysis
Contour Processing
This section describes contour processing functions.

Polygonal Approximation
As soon as all the borders have been retrieved from the image, the shape representation can be further compressed. Several algorithms are available for the purpose, including RLE coding of chain codes, higher order codes (see Figure 4-1), polygonal approximation, etc.
Figure 4-1 Higher Order Freeman Codes

24-Point Extended Chain Code

4-1

OpenCV Reference Manual

Structural Analysis

Polygonal approximation is the best method in terms of the output data simplicity for further processing. Below follow descriptions of two polygonal approximation algorithms. The main idea behind them is to find and keep only the dominant points, that is, points where the local maximums of curvature absolute value are located on the digital curve, stored in the chain code or in another direct representation format. The first step here is the introduction of a discrete analog of curvature. In the continuous case the curvature is determined as the speed of the tangent angle changing:
x y x y k = ---------------------------------- . 2 2 32 ( x + y )

In the discrete case different approximations are used. The simplest one, called L1 curvature, is the difference between successive chain codes:
ci
(1 )

= ( ( f i f i 1 + 4 ) mod 8 ) 4 .

(4.1)

This method covers the changes from 0, that corresponds to the straight line, to 4, that corresponds to the sharpest angle, when the direction is changed to reverse. The following algorithm is used for getting a more complex approximation. First, for the given point (xi, yi) the radius mi of the neighborhood to be considered is selected. For some algorithms mi is a method parameter and has a constant value for all points; for others it is calculated automatically for each point. The following value is calculated for all pairs (xi-k, yi-k) and (xi+k, yi+k) (k=1...m):
( a ik b ik ) c ik = ---------------------------- = cos ( a ik , b ik ) , a ik b ik

where

a ik = ( x i k x i , y i k y i ) , b ik = ( x i + k x i , yi k y i ) .

The next step is finding the index hi such that c im < c im 1 < < c ih i c ih i 1 . The value c ih is regarded as the curvature value of the ith point. The point value changes from i 1 (straight line) to 1 (sharpest angle). This approximation is called the k-cosine curvature. Rosenfeld-Johnston algorithm [Rosenfeld73] is one of the earliest algorithms for determining the dominant points on the digital curves. The algorithm requires the parameter m, the neighborhood radius that is often equal to 1/10 or 1/15 of the number of points in the input curve. Rosenfeld-Johnston algorithm is used to calculate curvature values for all points and remove points that satisfy the condition
j , i j hi 2 ; c ih < c jh
i j

4-2

OpenCV Reference Manual

Structural Analysis

The remaining points are treated as dominant points. Figure 4-2 shows an example of applying the algorithm.
Figure 4-2 Rosenfeld-Johnston Output for F-Letter Contour

Source Image

Rosenfeld-Johnston Algorithm Output

The disadvantage of the algorithm is the necessity to choose the parameter m and parameter identity for all the points, which results in either excessively rough, or excessively precise contour approximation. The next algorithm proposed by Teh and Chin [Teh89] includes a method for the automatic selection of the parameter m for each point. The algorithm makes several passes through the curve and deletes some points at each pass. At first, all points with zero c i ( 1 ) curvatures are deleted (see Equation 5.1). For other points the parameter mi and the curvature value are determined. After that the algorithm performs a non-maxima suppression, same as in Rosenfeld-Johnston algorithm, deleting points whose curvature satisfies the previous condition where for c i (1 ) the metric hi is set to mi. Finally, the algorithm replaces groups of two successive remaining points with a single point and groups of three or more successive points with a pair of the first and the last points. This algorithm does not require any parameters except for the curvature to use. Figure 4-3 shows the algorithm results.

4-3

OpenCV Reference Manual

Structural Analysis

Figure 4-3

Teh-Chin Output for F-Letter Contour

Source picture Source Picture

Teh-Chinalgorithm Output Teh-ChinAlgorithmoutput TC89 algorithm output

Douglas-Peucker Approximation
Instead of applying a rather sophisticated Teh-Chin algorithm to the chain code, the user may try another way to get a smooth contour on a little number of vertices. The idea is to apply some very simple approximation techniques to the chain code with polylines, such as substituting ending points for horizontal, vertical, and diagonal segments, and then use the approximation algorithm on polylines. This preprocessing reduces the amount of data without any accuracy loss. Teh-Chin algorithm also involves this step, but uses removed points for calculating curvatures of the remaining points. The algorithm to consider is a pure geometrical algorithm by Douglas-Peucker for approximating a polyline with another polyline with required accuracy: 1. Two points on the given polyline are selected, thus the polyline is approximated by the line connecting these two points. The algorithm iteratively adds new points to this initial approximation polyline until the

4-4

OpenCV Reference Manual

Structural Analysis

required accuracy is achieved. If the polyline is not closed, two ending points are selected. Otherwise, some initial algorithm should be applied to find two initial points. The more extreme the points are, the better. 2. The algorithm iterates through all polyline vertices between the two initial vertices and finds the farthest point from the line connecting two initial vertices. If this maximum distance is less than the required error, then the approximation has been found and the next segment, if any, is taken for approximation. Otherwise, the new point is added to the approximation polyline and the approximated segment is split at this point. Then the two parts are approximated in the same way, since the algorithm is recursive. For a closed polygon there are two polygonal segments to process.

Contours Moments
The moment of order (p; q) of an arbitrary region R is given by
pq =
x y dx dy .
p q

(4.2)

If p = q = 0 , we obtain the area a of R. The moments are usually normalized by the area a of R. These moments are called normalized moments:

Thus 00 interest:

= 1.

It is an explicit method for calculation of moments of arbitrary closed polygons. Contrary to most implementations that obtain moments from the discrete pixel data, this approach calculates moments by using only the border of a region. Since no explicit region needs to be constructed, and because the border of a region usually consists of significantly fewer points than the entire region, the approach is very efficient. The well-known Greens formula is used to calculate moments:

pq = 1 a

pq = ( 1 a )

x y dx dy .

(4.3)

For

p+q2
p

normalized central moments of R are usually the ones of

( x a 10 ) ( y a 01 ) dxdy

(4.4)

4-5

OpenCV Reference Manual

Structural Analysis

where b is the border of the region R. It follows from the formula (4.2) that:
Q x = x y , P y = 0 ,
p q

hence
P ( x , y ) = 0, Q ( x , y ) = 1 ( p + 1 ) x
p+1 q

Therefore, the moments from (4.2) can be calculated as follows:

If the border b consists of n points

b( t) =

where

b i ( t ) = tp + ( 1 t ) p i 1 .

Therefore, (4.5) can be calculated in the following manner: n

i = 1 bj

After unnormalized moments have been transformed, (4.6) could be written as:
1 v pA = -------------------------------------------------------------------------p+q (p + q + 2)(p + q + 1) p
n p q

i=1

k = 0i = 0

4-6

( xi 1 yi xiyi 1 )

v pq =

( 1 ( p + 1) x

p+1

y ) dy .

bi ( t ) ,
i=1

b i ( t ) , t [0,1]

is defined as

( 1 ( p + 1 )x

p+1

y ) dy

( Q ( x P y ) dx dy =

( P dx + Q dy ) ,

(4.5)
pi = ( xi , yi ) , 0 i n , p0 = pn ,

it follows that:

(4.6)

k+t p+qkt

x i xi 1 y i y i 1

k pk t qt

OpenCV Reference Manual

Structural Analysis

Central unnormalized and normalized moments up to order 3 look like n

i=1 n

i=1

2 xi ( y i 1

+ 3yi ) ) ,
n

i=1

yi( xi 1 + 3 xi ) ) ,
n

i=1

20 = 20 10 ,

a 03 = 1 ( 20 a )

a 12 = 1 ( 60 a )

a 21 = 1 ( 60 a )

a 30 = 1 ( 20 a )

a 02 = 1 ( 12 a )

a 11 = 1 ( 24 a )

a 20 = 1 ( 12 a )

a 01 = 1 ( 6 a )

a 10 = 1 ( 6 a )

a = 12

xi 1yi xi yi 1 ,

( xi 1 yi xiyi 1 ) ( xi 1 + xi ),

( xi 1 yi xiyi 1 ) ( yi 1 + yi ) ,

( x i 1 y i x i y i 1 ) ( xi 1 + x i 1 x i + x i ) ,

( x i 1 y i x i y i 1 ) ( 2 x i 1 + x i 1 yi + x i y i 1 + 2 x i y i ) ,

( x i 1 y i x i y i 1 ) ( yi 1 + y i 1 y i + y i ) ,

( x i 1 y i x i y i 1 ) ( xi 1 + x i 1 x i + x i x i 1 + x i ) ,

( x i 1 y i x i y i 1 ) ( xi 1 ( 3 y i 1 + y i ) + 2 x i 1 x i ( y i 1 + y i )

( x i 1 y i x i y i 1 ) ( yi 1 ( 3 x i 1 + x i ) + 2 y i 1 y i ( x i 1 + x i ) +

( x i 1 y i x i y i 1 ) ( yi 1 + y i 1 y i + y i y i 1 + y i ) ,

4-7

OpenCV Reference Manual

Structural Analysis

11 = 11 10 01 , 02 = 02 01 , 30 = 30 + 2 10 3 10 20 , 21 = 21 + 2 10 01 2 10 11 20 01 , 12 = 12 + 2 01 10 2 01 11 02 10 , 03 = 03 + 2 01 3 01 02 .
3 3 3 3 2

Hierarchical Representation of Contours

Let T be the simple closed boundary of a shape with n points T : { p ( 1 ) , p ( 2 ) , , p ( n ) } and n runs: { s ( 1 ) , s ( 2 ) , , s ( n ) }. Every run s ( i ) is formed by the two points ( p ( i ) , p ( i + 1 ) ) . For every pair of the neighboring runs s ( i ) and s ( i + 1 ) a triangle is defined by the two runs and the line connecting the two far ends of the two runs (Figure 4-4).
Figure 4-4 Triangles Numbering

p (i )

s(i )

p(i + 1)
s(i + 1)

t (i )

Triangles t ( i 2 ) , t ( i 1 ) , t ( i + 1 ) , t ( i + 2 ) are called neighboring triangles of (Figure 4-5).

Figure 4-7

Classification of Child Triangles

L child

R child

Structural Analysis

Cumulative tree distance shows the dissimilarity between the approximations of the two shapes and exhibits the multiresolution nature of the tree representation in shape matching. The shape matching algorithm is quite straightforward. For two given tree representations the two trees are traversed according to the breadth-first sequence to find CNPs of the two trees. Next dn[A(i),B(i)] and dc(TA,TB,i)are calculated for every i. If for some i dc(TA,TB,i)is larger than the tolerance threshold value, the matching procedure is terminated to indicate that the two shapes are dissimilar, otherwise it continues. If dt(TA,TB) is still less than the tolerance threshold value, then the procedure is terminated to indicate that there is a good match between TA and TB.

Geometry
This section describes functions from computational geometry field.

Ellipse Fitting
Fitting of primitive models to the image data is a basic task in pattern recognition and computer vision. A successful solution of this task results in reduction and simplification of the data for the benefit of higher level processing stages. One of the most commonly used models is the ellipse which, being a perspective projection of the circle, is of great importance for many industrial applications. The representation of general conic by the second order polynomial is T 2 2 F ( a , x ) = a , x = a x + bxy + cy + dx + ey + f = 0 with the vectors denoted as T T 2 2 a = [ a , b , c , d , e , f ] and x = [ x , xy , y , x , y , 1 ] .
F(a, x )

is called the algebraic distance between point

( x0 , y0 )
2

and conic

F ( a , x ) .

i=1

In order to achieve ellipse-specific fitting polynomial coefficients must be constrained. For ellipse they must satisfy b 2 4 ac < 0 .

4-14

Minimizing the sum of squared algebraic distances of conic.

F (x 0 )

may approach the fitting

OpenCV Reference Manual

Structural Analysis

Moreover, the equality constraint 4 ac b 2 coefficients scaling into constraint.

= 1 can
T

be imposed in order to incorporate

This constraint may be written as a matrix a

Ca = 1 . Da
2

Finally, the problem could be formulated as minimizing T T a Ca = 1 , where D is the nx6 matrix [ x 1 , x 2 , , x n ] . Introducing the Lagrange multiplier results in the system
2 D Da 2 Ca = 0 T a Ca = 1
Sa = 2 Ca a Ca = 1.
T T

with constraint

, which can be re-written as

The system solution is described in [Fitzgibbon95]. After the system is solved, ellipse center and axis can be extracted.

Line Fitting
M-estimators are used for approximating a set of points with geometrical primitives e.g., conic section, in cases when the classical least squares method fails. For example, the image of a line from the camera contains noisy data with many outliers, that is, the points that lie far from the main group, and the least squares method fails if applied. The least squares method searches for a parameter set that minimizes the sum of squared distances:

where d i is the distance from the ith point to the primitive. The distance type is specified as the function input parameter. If even a few points have a large di , then the perturbation in the primitive parameter values may be prohibitively big. The solution is to minimize

m =

i i

m =

di ,

( di ) ,

4-15

OpenCV Reference Manual

Structural Analysis

where ( d i ) grows slower than d 2 . This problem can be reduced to weighted least i squares [Fitzgibbon95], which is solved by iterative finding of the minimum of where k is the iteration number, d k 1 is the minimizer of the sum on the previous i 1 d iteration, and W ( x ) = --- ------ . If d i is a linear function of parameters p j d i = A ij p j x dx

then the minimization vector of the m k is the eigenvector of corresponds to the smallest eigenvalue. For more information see [Zhang96].

matrix that

Convexity Defects
Let ( p 1 , p 2 , p n ) be a closed simple polygon, or contour, and ( h1 , h 2 , h m ) a convex hull. A sequence of contour points exists normally between two consecutive convex hull vertices. This sequence forms the so-called convexity defect for which some useful characteristics can be computed. Computer Vision Library computes only one such characteristic, named depth (see Figure 4-10).
Figure 4-10 Convexity Defects

4-16

mk =

W ( di

)di ,

OpenCV Reference Manual

Structural Analysis

The black lines belong to the input contour. The red lines update the contour to its convex hull. The symbols s and e signify the start and the end points of the convexity defect. The symbol d is a contour point located between s and e being the farthermost from the line that includes the segment se. The symbol h stands for the convexity defect depth, that is, the distance from d to the se line. See CvConvexityDefect structure definition in Structural Analysis Reference.

4-17

OpenCV Reference Manual

Structural Analysis

4-18

Object Recognition
Eigen Objects
This section describes functions that operate on eigen objects.

Let us define an object u = { u 1, u 2 , u n } as a vector in the n-dimensional space. For example, u can be an image and its components ul are the image pixel values. In this case n is equal to the number of pixels in the image. Then, consider a group of input objects u i = { u i, u i, , u i } , where i = 1, , m and usually m << n. The averaged, or 1 2 n mean, object u = { u1, u2, , u n } of this group is defined as follows:
m

k=1

Covariance matrix C = |cij| is a square symmetric matrix n

l=1

Eigen objects basis e i = { e i, e i, , e i } , i = 1, , 1 2 n may be calculated using the following relation:

k=1

where i and of matrix C.

1 i e l = --------i

c ij =

1 u l = -m

ul .

i ( ul

ul )

j ( ul

ul ) .

m1 m

of the input objects group

vk ( ul ul ) ,

= { v 1 , v 2 , , v m }

are eigenvalues and the corresponding eigenvectors

5-1

OpenCV Reference Manual

Object Recognition

Any input object ui as well as any other object u may be decomposed in the eigen objectsnm1-D sub-space. Decomposition coefficients of the object u are:
l=1

Using these coefficients, we may calculate projection u = { u 1, u 2 , u n } of the object u to the eigen objects sub-space, or, in other words, restore the object u in that sub-space:
m1 k

k=1

For examples of use of the functions and relevant data types see Image Recognition Reference Chapter.

Embedded Hidden Markov Models

This section describes functions for using Embedded Hidden Markov Models (HMM) in face recognition task. See Reference for HMM Structures.

ul =

wi =

el ( ul ul ) .

wk el + ul .

5-2

OpenCV Reference Manual

Object Recognition

5-3

OpenCV Reference Manual

Object Recognition

5-4

3D Reconstruction
Camera Calibration
This section describes camera calibration and undistortion functions.

Camera Parameters
Camera calibration functions are used for calculating intrinsic and extrinsic camera parameters. Camera parameters are the numbers describing a particular camera configuration. The intrinsic camera parameters specify the camera characteristics proper; these parameters are:

focal length, that is, the distance between the camera lens and the image plane, location of the image center in pixel coordinates, effective pixel size, radial distortion coefficient of the lens.

The extrinsic camera parameters describe spatial relationship between the camera and the world; they are

rotation matrix, translation vector.

They specify the transformation between the camera and world reference frames. A usual pinhole camera is used. The relationship between a 3D point projection m is given by the formula
m = A [ Rt ] M , M

and its image

where

is the camera intrinsic matrix:

6-1

OpenCV Reference Manual

3D Reconstruction

fx 0 cx A =

0 fy cy 0 0 1

,where

( c x, c y ) ( f x, f y ) ( R, t )

are coordinates of the principal point; are the focal lengths by the axes x and y;
t

are extrinsic parameters: the rotation matrix R and translation vector relate the world coordinate system to the camera coordinate system:
r 11 r 12 r 13 R = r 21 r 22 r 23 r 31 r 32 r 33 t1

that

t = t2 t3

Camera usually exhibits significant lens distortion, especially radial distortion. The distortion is characterized by four coefficients: k1, k2, p1, p2. The functions UnDistortOnce and UnDistortInit + UnDistort correct the image from the camera given the four coefficients (see Figure 6-2). The following algorithm was used for camera calibration: 1. Find homography for all points on series of images. 2. Initialize intrinsic parameters; distortion is set to 0. 3. Find extrinsic parameters for each image of pattern. 4. Make main optimization by minimizing error of projection points with all parameters.

Homography
h 11 h 12 h 13 H = h 21 h 22 h 23 h 31 h 32 h 33

is the matrix of homography.

Without any loss of generality, the model plane may be assumed to be Z = 0 of the world coordinate system. If r i denotes the ith column of the rotation matrix R , then:

6-2

OpenCV Reference Manual

3D Reconstruction

X u X s v = A [r 1 r2 r3 t ] = Y = A [ r1 r2 t ] Y 0 l 1 1

By abuse of notation, M is still used to denote a point on the model plane, that is, T T M [ X, Y ] , since Z is always equal to 0. In its turn, M = [ X, Y, 1 ] . Therefore, a model point M and its image m are related by the homography H :
sm = HM

with

H = A [ r1 r2 t ] . H

It is clear that the 3x3 matrix

is defined without specifying a scalar factor.

Pattern
To calibrate the camera, the calibration routine is supplied with several views of a planar model object, or pattern, of known geometry. For every view the points on the model plane and their projections onto the image are passed to the calibration routine. In OpenCV a chessboard pattern is used (see Figure 6-1). To achieve more accurate calibration results, print out the pattern at high resolution on high-quality paper and put it on a hard, preferably glass, substrate.
Figure 6-1 Pattern

6-3

OpenCV Reference Manual

3D Reconstruction

Lens Distortion
Any camera usually exhibits significant lens distortion, especially radial distortion. The distortion is described by four coefficients: two radial distortion coefficients k1, k2, and two tangential ones p1, p2. Let ( u, v ) be true pixel image coordinates, that is, coordinates with ideal projection, and ( u, v ) be corresponding real observed (distorted) image coordinates. Similarly, ( x, y ) are ideal (distortion-free) and ( x, y ) are real (distorted) image physical coordinates. Taking into account two expansion terms gives the following:
x = x + x [ k 1 r + k 2 r ] + [ 2 p 1 xy + p 2 ( r + 2 x ) ] y = y + y [ k 1 r + k 2 r ] + [ 2 p 2 xy + p 2 ( r + 2 y ) ] ,
2 4 2 2 2 4 2 2

where r2 = x2 + y2. Second addends in the above relations describe radial distortion and the third ones - tangential. The center of the radial distortion is the same as the principal point. Because u = c x + f x u and v = c y + f y v , where cx, cy, fx, and fy are components of the camera intrinsic matrix, the resultant system can be rewritten as follows:

r 2 4 u = u + ( u c x ) k 1 r + k 2 r + 2 p1 y + p2 ----- + 2 x x

r 2 4 v = v + ( v c y ) k 1 r + k 2 r + 2 p 2 x + p 1 ----- + 2 y . y

The latter relations are used to undistort images from the camera. The group of camera undistortion functions consists of UnDistortOnce, UnDistortInit, and UnDistort. If only a single image is required to be corrected, cvUnDistortOnce function may be used. When dealing with a number of images possessing similar parameters, e.g., a sequence of video frames, use the other two functions. In this case the following sequence of actions must take place: 1. Allocate data array of length
<image_width>*<image_height>*<number_of_image_channels>.

2. Call the function UnDistortInit that fills the data array. 3. Call the function UnDistort for each frame from the camera.

6-4

OpenCV Reference Manual

3D Reconstruction

Figure 6-2

Correcting Lens Distortion

Im age W ith L ens D istortion

Im age W ith C orrected L ens D istortion

Rotation Matrix and Rotation Vector

Rodrigues conversion function Rodrigues is a method to convert a rotation vector to a rotation matrix or vice versa.

View Morphing
This section describes functions for morphing views from two cameras. The View Morphing technique is used to get an image from a virtual camera that could be placed between two real cameras. The input for View Morphing algorithms are two images from real cameras and information about correspondence between regions in the two images. The output of the algorithms is a synthesized image - "a view from the virtual camera". This section addresses the problem of synthesizing images of real scenes under three-dimensional transformation in viewpoint and appearance. Solving this problem enables interactive viewing of remote scenes on a computer, in which a user can move the virtual camera through the environment. A three-dimensional scene transformation can be rendered on a video display device through applying simple transformation to a

6-5

OpenCV Reference Manual

3D Reconstruction

set of basis images of the scene. The virtue of these transformations is that they operate directly on the image and recover only the scene information that is required to accomplish the desired effect. Consequently, the transformations are applicable in a situation when accurate three-dimensional models are difficult or impossible to obtain. The algorithm for synthesis of a virtual camera view from a pair of images taken from real cameras is shown below.

Algorithm
1. Find fundamental matrix, for example, using correspondence points in the images. 2. Find scanlines for each image. 3. Warp the images across the scanlines. 4. Find correspondence of the warped images. 5. Morph the warped images across position of the virtual camera. 6. Unwarp the image. 7. Delete moire from the resulting image.
Figure 6-3 Original Images

Original Image From Left Camera

Original Image From Right Camera

6-6

OpenCV Reference Manual

3D Reconstruction

Figure 6-4

Correspondence Points

Correspondence Points on Left Image

Correspondence Points on Right Image

Figure 6-5

Scan Lines

Some Scanlines on Left Iimage

Some Scanlines on Right Image

6-7

OpenCV Reference Manual

3D Reconstruction

Figure 6-6

Moire in Morphed Image

Figure 6-7

Resulting Morphed Image

Morphed Image From Virtual Camera With Deleted Moire

Using Functions for View Morphing Algorithm

1. Find the fundamental matrix using the correspondence points in the two images of cameras by calling the function FindFundamentalMatrix. 2. Find the number of scanlines in the images for the given fundamental matrix by calling the function FindFundamentalMatrix with null pointers to the scanlines.

6-8

OpenCV Reference Manual

Similarly, the projection of this point, namely p 0 , is a reference point for the image points: p i = p i p 0 . As follows from the weak-perspective assumption, the x component of p i is a scaled-down form of the x component of P i :
x i x0 = s ( Xi X 0 ) = s ( P 0 i ) .

(6.3)

This is also true for their y components. If I and J are defined as scaled-up versions of the unit vectors i and j ( I = si and J = sj ), then x i x0 = Pi I and y i y 0 = P i J (6.4) as two equations for each point for which I and J are unknown. These equations, collected over all the points, can be put into matrix form as
x = MI

and

y = MJ ,

(6.5)

where x and y are vectors of x and y components of p i respectively, and M is a matrix whose rows are the P i vectors. These two sets of equations can be further joined to construct a single set of linear equations:

[x y] = M[I J ]

piC = M [I J ] ,

(6.6)

6-12

OpenCV Reference Manual

3D Reconstruction

where p i is a matrix whose rows are p i . The latter equation is an overconstrained system of linear equations that can be solved for I and J in a least-squares sense as
[ I J ] = M pi ,
+

(6.7)
M.

where

is the pseudo-inverse of

Now that we have I and J , we construct the pose estimate as follows. First, i and j are estimated as I and J normalized, that is, scaled to unit length. By construction, these are the first two rows of the rotation matrix, and their cross-product is the third row: i
R =
T T T

(6.8)

(i j)

The average of the magnitudes of I and J is an estimate of the weak-perspective scale s . From the weak-perspective equations, the world point P 0 in camera coordinates is the image point p 0 in camera coordinates scaled by s:
P0 = p0 s = [ x0 y0 f ] s ,

(6.9)

which is precisely the translation vector being sought.

Algorithm
The POSIT algorithm was first presented in the paper by DeMenthon and Davis [DeMenthon92]. In this paper, the authors first describe their POS (Pose from Orthography and Scaling) algorithm. By approximating perspective projection with weak-perspective projection POS produces a pose estimate from a given image. POS can be repeatedly used by constructing a new weak perspective image from each pose estimate and feeding it into the next iteration. The calculated images are estimates of the initial perspective image with successively smaller amounts of perspective distortion so that the final image contains no such distortion. The authors term this iterative use of POS as POSIT (POS with ITerations). POSIT requires three pieces of known information:

6-13

OpenCV Reference Manual

3D Reconstruction

The object model, consisting of N points, each with unique 3D coordinates. N must
be greater than 3, and the points must be non-degenerate (non-coplanar) to avoid algorithmic difficulties. Better results are achieved by using more points and by choosing points as far from coplanarity as possible. The object model is an N x 3 matrix.

The object image, which is the set of 2D points resulting from a camera projection
of the model points onto an image plane; it is a function of the object current pose. The object image is an N x 2 matrix.

The camera intrinsic parameters, namely, the focal length of the camera.
Given the object model and the object image, the algorithm proceeds as follows: 1. The object image is assumed to be a weak perspective image of the object, from which a least-squares pose approximation is calculated via the object model pseudoinverse. 2. From this approximate pose the object model is projected onto the image plane to construct a new weak perspective image. 3. From this image a new approximate pose is found using least-squares, which in turn determines another weak perspective image, and so on. For well-behaved inputs, this procedure converges to an unchanging weak perspective image, whose corresponding pose is the final calculated object pose.
Example 6-1 POSIT Algorithm in Pseudo-Code

POSIT (imagePoints, objectPoints, focalLength) { count = converged = 0; modelVectors = modelPoints modelPoints(0); oldWeakImagePoints = imagePoints; while (!converged) { if (count == 0) imageVectors = imagePoints imagePoints(0); else { weakImagePoints = imagePoints .* ((1 + modelVectors*row3/translation(3)) * [1 1]); imageDifference = sum(sum(abs( round(weakImagePoints) round(oldWeakImagePoints)))); oldWeakImagePoints = weakImagePoints; imageVectors = weakImagePoints weakImagePoints(0); } [I J] = pseudoinverse(modelVectors) * imageVectors;

calculate the homography matrix for the initial image transformation from the
image plane to the plane defined by the frontal arm plane (the function CalcImageHomography),

calculate the probability density histogram for the arm location (the function
CalcProbDensity).

6-18

Basic Structures and Operations

Image Functions
This section describes basic functions for manipulating raster images.

OpenCV library represents images in the format IplImage that comes from Intel Image Processing Library (IPL). IPL reference manual gives detailed information about the format, but, for completeness, it is also briefly described here.
Example 7-1
IplImage Structure Definition typedef struct _IplImage { int nSize; /* size of iplImage struct */ int ID; /* image header version */ int nChannels; int alphaChannel; int depth; /* pixel depth in bits */ char colorModel[4]; char channelSeq[4]; int dataOrder; int origin; int align; /* 4- or 8-byte align */ int width; int height; struct _IplROI *roi; /* pointer to ROI if any */ struct _IplImage *maskROI; /*pointer to mask ROI if any */ void *imageId; /* use of the application */ struct _IplTileInfo *tileInfo; /* contains information on tiling */ int imageSize; /* useful size in bytes */ char *imageData; /* pointer to aligned image */ int widthStep; /* size of aligned line in bytes */ int BorderMode[4]; /* the top, bottom, left, and right border mode */ int BorderConst[4]; /* constants for the top, bottom, left, and right border */ char *imageDataOrigin; /* ptr to full, nonaligned image */ } IplImage;

7-1

OpenCV Reference Manual

Basic Structures and Operations

Only a few of the most important fields of the structure are described here. The fields width and height contain image width and height in pixels, respectively. The field depth contains information about the type of pixel values. All possible values of the field depth listed in ipl.h header file include:
IPL_DEPTH_8U IPL_DEPTH_8S -

unsigned 8-bit integer value (unsigned char), signed 8-bit integer value (signed char or simply char), signed 16-bit integer value (short int),

IPL_DEPTH_16S IPL_DEPTH_32S

- signed 32-bit integer value (int), 32-bit floating-point single-precision value (float).

IPL_DEPTH_32F -

In the above list the corresponding types in C are placed in parentheses. The parameter
nChannels means the number of color planes in the image. Grayscale images contain a

single channel, while color images usually include three or four channels. The parameter origin indicates, whether the top image row (origin == IPL_ORIGIN_TL) or bottom image row (origin == IPL_ORIGIN_BL) goes first in memory. Windows bitmaps are usually bottom-origin, while in most of other environments images are top-origin. The parameter dataOrder indicates, whether the color planes in the color image are interleaved (dataOrder == IPL_DATA_ORDER_PIXEL) or separate (dataOrder == IPL_DATA_ORDER_PLANE). The parameter widthStep contains the number of bytes between points in the same column and successive rows. The parameter width is not sufficient to calculate the distance, because each row may be aligned with a certain number of bytes to achieve faster processing of the image, so there can be some gaps between the end of ith row and the start of (i+1)th row. The parameter imageData contains pointer to the first row of image data. If there are several separate planes in the image (when dataOrder == IPL_DATA_ORDER_PLANE), they are placed consecutively as separate images with height*nChannels rows total.

7-2

OpenCV Reference Manual

Basic Structures and Operations

It is possible to select some rectangular part of the image or a certain color plane in the image, or both, and process only this part. The selected rectangle is called "Region of Interest" or ROI. The structure IplImage contains the field roi for this purpose. If the pointer not NULL, it points to the structure IplROI that contains parameters of selected ROI, otherwise a whole image is considered selected.
Example 7-2
IplROI Structure Definition

typedef struct _IplROI { int coi; /* channel of interest or COI */ int xOffset; int yOffset; int width; int height; } IplROI;

As can be seen, IplROI includes ROI origin and size as well as COI (Channel of Interest) specification. The field coi, equal to 0, means that all the image channels are selected, otherwise it specifies an index of the selected image plane. Unlike IPL, OpenCV has several limitations in support of IplImage: Each function supports only a few certain depths and/or number of channels. For example, image statistics functions support only single-channel or three-channel images of the depth IPL_DEPTH_8U, IPL_DEPTH_8S or IPL_DEPTH_32F. The exact information about supported image formats is usually contained in the description of parameters or in the beginning of the chapter if all the functions described in the chapter are similar. It is quite different from IPL that tries to support all possible image formats in each function. OpenCV supports only interleaved images, not planar ones. The fields colorModel, channelSeq, BorderMode, and BorderConst are ignored. The field align is ignored and widthStep is simply used instead of recalculating it using the fields width and align. The fields maskROI and tileInfo must be zero. COI support is very limited. Now only image statistics functions accept non-zero COI values. Use the functions CvtPixToPlane and CvtPlaneToPix as a work-around.

7-3

OpenCV Reference Manual

Basic Structures and Operations

ROIs of all the input/output images have to match exactly one another. For example, input and output images of the function Erode must have ROIs with equal sizes. It is unlike IPL again, where the ROIs intersection is actually affected. Despite all the limitations, OpenCV still supports most of the commonly used image formats that can be supported by IplImage and, thus, can be successfully used with IPL on common subset of possible IplImage formats. The functions described in this chapter are mainly short-cuts for operations of creating, destroying, and other common operations on IplImage, and they are often implemented as wrappers for original IPL functions.

Dynamic Data Structures

This chapter describes several resizable data structures and basic functions that are designed to operate on these structures.

Memory Storage
Memory storages provide the space for storing all the dynamic data structures described in this chapter. A storage consists of a header and a double-linked list of memory blocks. This list is treated as a stack, that is, the storage header contains a pointer to the block that is not occupied entirely and an integer value, the number of free bytes in this block. When the free space in the block has run out, the pointer is moved to the next block, if any, otherwise, a new block is allocated and then added to the list of blocks. All the blocks are of the same size and, therefore, this technique ensures an accurate memory allocation and helps avoid memory fragmentation if the blocks are large enough (see Figure 7-1).

7-4

OpenCV Reference Manual

Basic Structures and Operations

Figure 7-1

Memory Storage Organization

Storage Header BOTTOM TOP Free Space

Memory Blocks

Sequences
A sequence is a resizable array of arbitrary type elements located in the memory storage. The sequence is discontinuous. Sequence data may be partitioned into several continuous blocks, called sequence blocks, that can be located in different memory blocks. Sequence blocks are connected into a circular double-linked list to store large sequences in several memory blocks or keep several small sequences in a single memory block. For example, such organization is suitable for storing contours. The sequence implementation provides fast functions for adding/removing elements to/from the head and tail of the sequence, so that the sequence implements a deque. The functions for inserting/removing elements in the middle of a sequence are also available but they are slower. The sequence is the basic type for many other dynamic data structures in the library, e.g., sets, graphs, and contours; just like all these types, the sequence never returns the occupied memory to the storage. However, the sequence keeps track of the memory released after removing elements from the

7-5

OpenCV Reference Manual

Basic Structures and Operations

sequence; this memory is used repeatedly. To return the memory to the storage, the user may clear a whole storage, or use save/restoring position functions, or keep temporary data in child storages.
Figure 7-2 Sequence Structure
Storage Header Links Between Blocks.

Sequence Header and, probably, the First Sequence Block.

Sequence Blocks.

Writing and Reading Sequences

Although the functions and macros described below are irrelevant in theory because functions like SeqPush and GetSeqElem enable the user to write to sequences and read from them, the writing/reading functions and macros are very useful in practice because of their speed. The following problem could provide an illustrative example. If the task is to create a function that forms a sequence from N random values, the PUSH version runs as follows:
CvSeq* create_seq1( CvStorage* storage, int N ) { CvSeq* seq = cvCreateSeq( 0, sizeof(*seq), sizeof(int), storage); for( int i = 0; i < N; i++ ) { int a = rand(); cvSeqPush( seq, &a ); } return seq;

7-6

OpenCV Reference Manual

Basic Structures and Operations

The second version makes use of the fast writing scheme, that includes the following steps: initialization of the writing process (creating writer), writing, closing the writer (flush).
CvSeq* create_seq1( CvStorage* storage, int N ) { CvSeqWriter writer; cvStartWriteSeq( 0, sizeof(*seq), sizeof(int), storage, &writer ); for( int i = 0; i < N; i++ ) { int a = rand(); CV_WRITE_SEQ_ELEM( a, writer ); } return cvEndWriteSeq( &writer ); }

If N = 100000 and Pentium III 500MHz is used, the first version takes 230 milliseconds and the second one takes 111 milliseconds to finish. These characteristics assume that the storage already contains a sufficient number of blocks so that no new blocks are allocated. A comparison with the simple loop that does not use sequences gives an idea as to how effective and efficient this approach is.
int* create_seq3( int* buffer, int N ) { for( i = 0; i < N; i++ ) { buffer[i] = rand(); } return buffer; }

This function takes 104 milliseconds to finish using the same machine. Generally, the sequences do not make a great impact on the performance and the difference is very insignificant (less than 7% in the above example). However, the advantage of sequences is that the user can operate the input or output data even without knowing their amount in advance. These structures enable him/her to allocate memory iteratively. Another problem solution would be to use lists, yet the sequences are much faster and require less memory.

7-7

OpenCV Reference Manual

Basic Structures and Operations

Sets
The set structure is mostly based on sequences but has a totally different purpose. For example, the user is unable to use sequences for location of the dynamic structure elements that have links between one another because if some elements have been removed from the middle of the sequence, other sequence elements are moved to another location and their addresses and indices change. In this case all links have to be fixed anew. Another aspect of this problem is that removing elements from the middle of the sequence is slow, with time complexity of O(n), where n is the number of elements in the sequence. The problem solution lies in making the structure sparse and unordered, that is, whenever a structure element is removed, other elements must stay where they have been, while the cell previously occupied by the element is added to the pool of three cells; when a new element is inserted into the structure, the vacant cell is used to store this new element. The set operates in this way (See Example 7-3). The set looks like a list yet keeps no links between the structure elements. However, the user is free to make and keep such lists, if needed. The set is implemented as a sequence subclass; the set uses sequence elements as cells and organizes a list of free cells.

7-8

OpenCV Reference Manual

Basic Structures and Operations

See Figure 7-3 for an example of a set. For simplicity, the figure does not show division of the sequence/set into memory blocks and sequence blocks.
Figure 7-3 Set Structure

Existing Set Elements List of Free Cells

Set Header

Free Cells, Linked Together

The set elements, both existing and free cells, are all sequence elements. A special bit indicates whether the set element exists or not: in the above diagram the bits marked by 1 are free cells and the ones marked by 0 are occupied cells. The macro CV_IS_SET_ELEM_EXISTS(set_elem_ptr) uses this special bit to return a non-zero value if the set element specified by the parameter set_elem_ptr belongs to the set, and 0 otherwise. Below follows the definition of the structure CvSet:
Example 7-3
CvSet Structure Definition \ \

#define CV_SET_FIELDS() CV_SEQUENCE_FIELDS() CvMemBlock* free_elems; typedef struct CvSet { CV_SET_FIELDS() } CvSet;

In other words, a set is a sequence plus a list of free cells.

7-9

OpenCV Reference Manual

Basic Structures and Operations

There are two modes of working with sets: 1. Using indices for referencing the set elements within a sequence 2. Using pointers for the same purpose. Whereas at times the first mode is a better option, the pointer mode is faster because it does not need to find the set elements by their indices, which is done in the same way as in simple sequences. The decision on which method should be used in each particular case depends on:

the type of operations to be performed on the set and the way the operations on the set should be performed.
The ways in which a new set is created and new elements are added to the existing set are the same in either mode, the only difference between the two being the way the elements are removed from the set. The user may even use both methods of access simultaneously, provided he or she has enough memory available to store both the index and the pointer to each element. Like in sequences, the user may create a set with elements of arbitrary type and specify any size of the header subject to the following restrictions:

size of the header may not be less than sizeof(CvSet). size of the set elements should be divisible by 4 and not less than 8 bytes.
The reason behind the latter restriction is the internal set organization: if the set has a free cell available, the first 4-byte field of this set element is used as a pointer to the next free cell, which enables the user to keep track of all free cells. The second 4-byte field of the cell contains the cell to be returned when the cell becomes occupied. When the user removes a set element while operating in the index mode, the index of the removed element is passed and stored in the released cell again. The bit indicating whether the element belongs to the set is the least significant bit of the first 4-byte field. This is the reason why all the elements must have their size divisible by 4. In this case they are all aligned with the 4-byte boundary, so that the least significant bits of their addresses are always 0. In free cells the corresponding bit is set to 1 and, in order to get the real address of the next free cell, the functions mask this bit off. On the other hand, if the cell is occupied, the corresponding bit must be equal to 0, which is the second and last restriction: the

7-10

OpenCV Reference Manual

Basic Structures and Operations

least significant bit of the first 4-byte field of the set element must be 0, otherwise the corresponding cell is considered free. If the set elements comply with this restriction, e.g., if the first field of the set element is a pointer to another set element or to some aligned structure outside the set, then the only restriction left is a non-zero number of 4- or 8-byte fields after the pointer. If the set elements do not comply with this restriction, e.g., if the user wants to store integers in the set, the user may derive his or her own structure from the structure CvSetElem or include it into his or her structure as the first field.
Example 7-4
CvSetElem Structure Definition \

#define CV_SET_ELEM_FIELDS() int* aligned_ptr; typedef struct _CvSetElem { CV_SET_ELEM_FIELDS() } CvSetElem;

The first field is a dummy field and is not used in the occupied cells, except the least significant bit, which is 0. With this structure the integer element could be defined as follows:
typedef struct _IntSetElem { CV_SET_ELEM_FIELDS() int value; } IntSetElem;

Graphs
The structure set described above helps to build graphs because a graph consists of two sets, namely, vertices and edges, that refer to each other.
Example 7-5
CvGraph Structure Definition \ \

#define CV_GRAPH_FIELDS() CV_SET_FIELDS() CvSet* edges; typedef struct _CvGraph

7-11

OpenCV Reference Manual

Basic Structures and Operations

Example 7-5
{

CvGraph Structure Definition (continued)

CV_GRAPH_FIELDS() } CvGraph;

In OOP terms, the graph structure is derived from the set of vertices and includes a set of edges. Besides, special data types exist for graph vertices and graph edges.
Example 7-6 Definitions of CvGraphEdge and CvGraphVtx Structures

#define CV_GRAPH_EDGE_FIELDS() \ struct _CvGraphEdge* next[2]; \ struct _CvGraphVertex* vtx[2]; #define CV_GRAPH_VERTEX_FIELDS() struct _CvGraphEdge* first; typedef struct _CvGraphEdge { CV_GRAPH_EDGE_FIELDS() } CvGraphEdge; typedef struct _CvGraphVertex { CV_GRAPH_VERTEX_FIELDS() } CvGraphVtx; \

The graph vertex has a single predefined field that assumes the value of 1 when pointing to the first edge incident to the vertex, or 0 if the vertex is isolated. The edges incident to a vertex make up the single linked non-cycle list. The edge structure is more complex: vtx [ 0 ] and vtx [ 1 ] are the starting and ending vertices of the edge, next[0] and next[1] are the next edges in the incident lists for vtx [ 0 ] and vtx [ 1 ]

7-12

OpenCV Reference Manual

Basic Structures and Operations

respectively. In other words, each edge is included in two incident lists since any edge is incident to both the starting and the ending vertices. For example, consider the following oriented graph (see below for more information on non-oriented graphs).
Figure 7-4 Sample Graph

0 1

3 2

The structure can be created with the following code:

CvGraph* graph = cvCreateGraph( CV_SEQ_KIND_GRAPH | CV_GRAPH_FLAG_ORIENTED, sizeof(CvGraph), sizeof(CvGraphVtx)+4, sizeof(CvGraphEdge), storage); for( i = 0; i < 5; i++ ) { cvGraphAddVtx( graph, 0, 0 );/* arguments like in cvSetAdd*/ } cvGraphAddEdge( graph, 0, 1, 0, 0 ); /* connect vertices 0

7-13

OpenCV Reference Manual

Basic Structures and Operations

and 1, other two arguments like in cvSetAdd */ cvGraphAddEdge( graph, 1, 2, 0, 0 ); cvGraphAddEdge( graph, 2, 0, 0, 0 ); cvGraphAddEdge( graph, 2, 3, 0, 0 );

The internal structure comes to be as follows:

Figure 7-5 Internal Structure for Sample Graph Shown in Figure 7-4

Graph vertices Graph Vertices Graph vertices 0 0 0 1 1 1 2 2 2 4 4 4 5 5 5

Graph Edges Graph edges Graph edges

Undirected graphs can also be represented by the structure CvGraph. If the non-oriented edges are substituted for the oriented ones, the internal structure remains the same. However, the function used to find edges succeeds only when it finds the edge from 3 to 2, as the function looks not only for edges from 3 to 2 but also from 2 to 3, and such an edge is present as well. As follows from the code, the type of the graph is specified when the graph is created, and the user can change the behavior of the edge searching function by specifying or omitting the flag CV_GRAPH_FLAG_ORIENTED. Two edges connecting the same vertices in undirected graphs may never be created because the existence of the edge between two vertices is checked before a new edge is inserted

7-14

OpenCV Reference Manual

Basic Structures and Operations

between them. However, internally the edge can be coded from the first vertex to the second or vice versa. Like in sets, the user may work with either indices or pointers. The graph implementation uses only pointers to refer to edges, but the user can choose indices or pointers for referencing vertices.

Matrix Operations
OpenCV introduces special data type CvMat for storing real single-precision or double-precision matrices. Operations supported include basic matrix arithmetics, eigen problem solution, SVD, 3D geometry and recognition-specific functions. To reduce time call overhead, the special data type CvMatArray, which is an array of matrices, and support functions are also introduced.

Library Technical Organization and System Functions

8-3

OpenCV Reference Manual

Library Technical Organization and System Functions

8-4

Motion Analysis and Object Tracking Reference

Table 9-1

Motion Analysis and Object Tracking Functions and Data Types

Group Background Subtraction Functions Name Functions Description Adds a new image to the accumulating sum. Calculates square of the source image and adds it to the destination image. Calculates product of two input images and adds it to the destination image. Calculates weighted sum of two images. Updates the motion history image by moving the silhouette. Calculates gradient orientation of the motion history image. Calculates the general motion direction in the selected region.

Acc SquareAcc

MultiplyAcc

RunningAvg
Motion Templates Functions

UpdateMotionHistory

CalcMotionGradient

CalcGlobalOrientation

9-1

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

Table 9-1

Motion Analysis and Object Tracking Functions and Data Types (continued)
Group Name Description Segments the whole motion into separate moving parts. Finds an object center using the MeanShift algorithm, calculates the object size and orientation. Iterates to find the object center. Changes contour position to minimize its energy. Calculates optical flow for two images implementing Horn and Schunk technique. Calculates optical flow for two images implementing Lucas and Kanade technique. Calculates optical flow for two images implementing the Block Matching algorithm. Calculates optical flow for two images using iterative Lucas-Kanade method in pyramids. Allocates Kalman filter structure. Deallocates Kalman filter structure. Estimates the subsequent stochastic model state.

SegmentMotion

CamShift Functions

CamShift

MeanShift
Active Contours Function

SnakeImage

Optical Flow Functions

CalcOpticalFlowHS

CalcOpticalFlowLK

CalcOpticalFlowBM

CalcOpticalFlowPyrLK

Estimators Functions

CreateKalman ReleaseKalman KalmanUpdateByTime

9-2

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

Table 9-1

Motion Analysis and Object Tracking Functions and Data Types (continued)
Group Name Description Adjusts the stochastic model state on the basis of the true measurements. Allocates a ConDensation filter structure. Deallocates a ConDensation filter structure. Initializes a sample set for condensation algorithm. Estimates the subsequent model state by its current state.

KalmanUpdateByMeasurement

CreateConDensation

ReleaseConDensation

ConDensInitSampleSet

ConDensUpdatebyTime

Data Types Estimators Data Types

CvKalman CvConDensation

Background Subtraction Functions

Acc
Adds frame to accumulator.
void cvAcc( IplImage* img, IplImage* sum, IplImage* mask=0 ); img sum mask

Input image. Accumulating image. Mask image.

9-3

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

Discussion The function Acc adds a new image img to the accumulating sum sum. If mask is not NULL, it specifies what accumulator pixels are affected.

SquareAcc
Calculates square of source image and adds it to destination image.
void cvSquareAcc( IplImage* img, IplImage* sqSum, IplImage* mask=0 ); img sqSum mask

Input image. Accumulating image. Mask image.

Discussion The function SquareAcc adds the square of the new image img to the accumulating sum sqSum of the image squares. If mask is not NULL, it specifies what accumulator pixels are affected.

MultiplyAcc
Calculates product of two input images and adds it to destination image.
void cvMultiplyAcc( IplImage* imgA, IplImage* imgB, IplImage* acc, IplImage* mask=0); imgA imgB acc

First input image. Second input image. Accumulating image.

9-4

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

mask

Mask image.

Discussion The function MultiplyAcc multiplies input imgA by imgB and adds the result to the accumulating sum acc of the image products. If mask is not NULL, it specifies what accumulator pixels are affected.

RunningAvg
Calculates weighted sum of two images.
void cvRunningAvg( IplImage* imgY, IplImage* imgU, double alpha, IplImage* mask=0 ) imgY imgU alpha mask

Input image. Destination image. Weight of input image. Mask image.

Discussion The function RunningAvg calculates weighted sum of two images. Once a statistical model is available, slow updating of the value is often required to account for slowly changing lighting, etc. This can be done by using a simple adaptive filter:
t = y + ( 1 ) t 1 ,

where (imgU) is the updated value, 0 1 is an averaging constant, typically set to a small value such as 0.05, and y (imgY) is a new observation at time t. When the function is applied to a frame sequence, the result is called the running average of the sequence. If mask is not NULL, it specifies what accumulator pixels are affected.

9-5

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

Motion Templates Functions

UpdateMotionHistory
Updates motion history image by moving silhouette.
void cvUpdateMotionHistory (IplImage* silhouette, IplImage* mhi, double timestamp, double mhiDuration); silhouette mhi timestamp mhiDuration

Silhouette image that has non-zero pixels where the motion occurs. Motion history image, both an input and output parameter. Floating point current time in milliseconds. Maximal duration of motion track in milliseconds.

Discussion The function UpdateMotionHistory updates the motion history image with a silhouette, assigning the current timestamp value to those mhi pixels that have corresponding non-zero silhouette pixels. The function also clears mhi pixels older than timestamp mhiDuration if the corresponding silhouette values are 0.

CalcMotionGradient
Calculates gradient orientation of motion history image.
void cvCalcMotionGradient( IplImage* mhi, IplImage* mask, IplImage* orientation, double maxTDelta, double minTDelta, int apertureSize=3 ); mhi mask

Motion history image. Mask image; marks pixels where motion gradient data is correct. Output parameter.

9-6

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

orientation

Motion gradient orientation image; contains angles from 0 to ~360 degrees. Size of aperture used to calculate derivatives. Value should be odd, e.g., 3, 5, etc. Upper threshold. The function considers the gradient orientation valid if the difference between the maximum and minimum mhi values within a pixel neighborhood is lower than this threshold. Lower threshold. The function considers the gradient orientation valid if the difference between the maximum and minimum mhi values within a pixel neighborhood is greater than this threshold.

apertureSize

maxTDelta

minTDelta

Discussion The function CalcMotionGradient calculates the derivatives Dx and Dy for the image mhi and then calculates orientation of the gradient using the formula
=
0,
x = 0, y = o

arc tan ( y x ) else

Finally, the function masks off pixels with a very small (less than minTDelta) or very large (greater than maxTDelta) difference between the minimum and maximum mhi values in their neighborhood. The neighborhood for determining the minimum and maximum has the same size as aperture for derivative kernels - apertureSize x apertureSize pixels.

CalcGlobalOrientation
Calculates global motion orientation of some selected region.
void cvCalcGlobalOrientation( IplImage* orientation, IplImage* mask, IplImage* mhi, double currTimestamp, double mhiDuration );

9-7

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

orientation

Motion gradient orientation image; calculated by the function CalcMotionGradient. Mask image. It is a conjunction of valid gradient mask, calculated by the function CalcMotionGradient and mask of the region, whose direction needs to be calculated. Motion history image. Current time in milliseconds. Maximal duration of motion track in milliseconds.

mask

mhi currTimestamp mhiDuration

Discussion The function CalcGlobalOrientation calculates the general motion direction in the selected region. At first the function builds the orientation histogram and finds the basic orientation as a coordinate of the histogram maximum. After that the function calculates the shift relative to the basic orientation as a weighted sum of all orientation vectors: the more recent is the motion, the greater is the weight. The resultant angle is <basic orientation> + <shift>.

SegmentMotion
Segments whole motion into separate moving parts.
void cvSegmentMotion( IplImage* mhi, IplImage* segMask, CvMemStorage* storage, CvSeq** components, double timestamp, double segThresh ); mhi segMask Storage

Motion history image. Image where the mask found should be stored. Pointer to the memory storage, where the sequence of components should be saved. Sequence of components found by the function. Floating point current time in milliseconds.

components timestamp

9-8

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

segThresh

Segmentation threshold; recommended to be equal to the interval between motion history steps or greater.

Discussion The function SegmentMotion finds all the motion segments, starting from connected components in the image mhi that have value of the current timestamp. Each of the resulting segments is marked with an individual value (1,2 ...). The function stores information about each resulting motion segment in the structure CvConnectedComp (See Example 10-1 in Image Analysis Reference). The function returns a sequence of such structures.

CamShift Functions

CamShift
Finds object center, size, and orientation.
int cvCamShift( IplImage* imgProb, CvRect windowIn, CvTermCriteria criteria, CvConnectedComp* out, CvBox2D* box=0 ); imgProb windowIn criteria

2D object probability distribution. Initial search window. Criteria applied to determine when the window search should be finished. Resultant structure that contains converged search window coordinates (rect field) and sum of all pixels inside the window (area field). Circumscribed box for the object. If not NULL, contains object size and orientation.

out

box

9-9

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

Discussion The function CamShift finds an object center using the Mean Shift algorithm and, after that, calculates the object size and orientation. The function returns number of iterations made within the Mean Shift algorithm.

MeanShift
Iterates to find object center.
int cvMeanShift( IplImage* imgProb, CvRect criteria, CvConnectedComp* out ); imgProb windowIn criteria windowIn, CvTermCriteria

out

Discussion The function MeanShift iterates to find the object center given its 2D color probability distribution image. The iterations are made until the search window center moves by less than the given value and/or until the function has done the maximum number of iterations. The function returns the number of iterations made.

9-10

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

Active Contours Function

SnakeImage
Changes contour position to minimize its energy.
void cvSnakeImage( IplImage* image, CvPoint* points, int length, float* alpha, float* beta, float* gamma, int coeffUsage,CvSize win, CvTermCriteria criteria, int calcGradient=1 ); image points length alpha beta gamma coeffUsage

Pointer to the source image. Points of the contour. Number of points in the contour. Weight of continuity energy. Weight of curvature energy. Weight of image energy. Variant of usage of the previous three parameters:

CV_VALUE indicates CV_ARRAY indicates

that each of alpha, beta, gamma is a pointer to a single value to be used for all points; that each of alpha, beta, gamma is a pointer to an array of coefficients different for all the points of the snake. All the arrays must have the size equal to the snake size.

win

Size of neighborhood of every point used to search the minimum; must be odd. Termination criteria. Gradient flag. If not 0, the function counts source image gradient magnitude as external energy, otherwise the image intensity is considered.

criteria calcGradient

Discussion The function SnakeImage uses image intensity as image energy.

9-11

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

The parameter criteria.epsilon is used to define the minimal number of points that must be moved during any iteration to keep the iteration process running. If the number of moved points is less than criteria.epsilon or the function performed criteria.maxIter iterations, the function terminates.

Optical Flow Functions

CalcOpticalFlowHS
Calculates optical flow for two images.
void cvCalcOpticalFlowHS( IplImage* imgA, IplImage* imgB, int usePrevious, IplImage* velx, IplImage* vely, double lambda, CvTermCriteria criteria); imgA imgB usePrevious velx vely lambda criteria

First image. Second image. Uses previous (input) velocity field. Horizontal component of the optical flow. Vertical component of the optical flow. Lagrangian multiplier. Criteria of termination of velocity computing.

Discussion The function CalcOpticalFlowHS computes flow for every pixel, thus output images must have the same size as the input. Horn & Schunck Technique is implemented.

9-12

CV_LKFLOW_PYR_B_READY, pyramid

CV_LKFLOW_PYR_A_READY,

for the second frame is

precalculated before the call;

CV_LKFLOW_INITIAL_GUESSES,

array B contains initial coordinates of features before the function call.

The function CalcOpticalFlowPyrLK calculates the optical flow between two images for the given set of points. The function finds the flow with sub-pixel accuracy. Both parameters pyrA and pyrB comply with the following rules: if the image pointer is 0, the function allocates the buffer internally, calculates the pyramid, and releases the buffer after processing. Otherwise, the function calculates the pyramid and stores it in the buffer unless the flag CV_LKFLOW_PYR_A[B]_READY is set. The image should be large enough to fit the Gaussian pyramid data. After the function call both pyramids are calculated and the ready flag for the corresponding image can be set in the next call.

9-15

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

Estimators Functions

CreateKalman
Allocates Kalman filter structure.
CvKalman* cvCreateKalman( int DynamParams, int MeasureParams ); DynamParams MeasureParams

Dimension of the state vector. Dimension of the measurement vector.

Discussion The function CreateKalman creates CvKalman structure and returns pointer to the structure.

ReleaseKalman
Deallocates Kalman filter structure.
void cvReleaseKalman(CvKalman** Kalman); Kalman

Double pointer to the structure to be released.

Discussion The function ReleaseKalman releases the structure CvKalman (see Example 9-1) and frees the memory previously allocated for the structure.

9-16

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

KalmanUpdateByTime
Estimates subsequent model state.
void cvKalmanUpdateByTime (CvKalman* Kalman); Kalman

Pointer to the structure to be updated.

Discussion The function KalmanUpdateByTime estimates the subsequent stochastic model state by its current state.

KalmanUpdateByMeasurement
Adjusts model state.
void cvKalmanUpdateByMeasurement (CvKalman* Kalman,CvMat* Measurement); Kalman Measurement

Pointer to the structure to be updated. Pointer to the structure CvMat containing the measurement vector.

Discussion The function KalmanUpdateByMeasurement adjusts stochastic model state on the basis of the true measurements of the model state.

CreateConDensation
Allocates ConDensation filter structure.
CvConDensation* cvCreateConDensation( int DynamParams, int MeasureParams, int SamplesNum);

9-17

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

DynamParams MeasureParams SamplesNum

Dimension of the state vector. Dimension of the measurement vector. Number of samples.

Discussion The function CreateConDensation creates cvConDensation (see Example 9-2) structure and returns pointer to the structure.

ReleaseConDensation
Deallocates ConDensation filter structure.
void cvReleaseConDensation(CvConDensation** ConDens); ConDens

Pointer to the pointer to the structure to be released.

Discussion The function ReleaseConDensation releases the structure CvConDensation (see Example 9-2) and frees all memory previously allocated for the structure.

ConDensInitSampleSet
Initializes sample set for condensation algorithm.
void cvConDensInitSampleSet(CvConDensation* ConDens, CvMat* lowerBound CvMat* upperBound); ConDens lowerBound upperBound

Pointer to a structure to be initialized. Vector of the lower boundary for each dimension. Vector of the upper boundary for each dimension.

9-18

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

Discussion The function ConDensInitSampleSet fills the samples arrays in the structure CvConDensation (see Example 9-2) with values within specified ranges.

ConDensUpdatebyTime
Estimates subsequent model state.
void cvConDensUpdateByTime(CvConDensation* ConDens); ConDens

Pointer to the structure to be updated.

Discussion The function ConDensUpdateByTime estimates the subsequent stochastic model state from its current state.

Estimators Data Types

Example 9-1
CvKalman

typedef struct CvKalman { int MP; //Dimension of measurement vector int DP; // Dimension of state vector float* PosterState; // Vector of State of the System in k-th step float* PriorState; // Vector of State of the System in (k-1)-th step float* DynamMatr; // Matrix of the linear Dynamics system float* MeasurementMatr; // Matrix of linear measurement float* MNCovariance; // Matrix of measurement noice covariance float* PNCovariance; // Matrix of process noice covariance float* KalmGainMatr; // Kalman Gain Matrix float* PriorErrorCovariance; //Prior Error Covariance matrix float* PosterErrorCovariance;//Poster Error Covariance matrix float* Temp1; // Temporary Matrixes float* Temp2; }CvKalman;

9-19

OpenCV Reference Manual

Motion Analysis and Object Tracking Reference

Example 9-2

CvConDensation

typedef struct { int MP; //Dimension of measurement vector int DP; // Dimension of state vector float* DynamMatr; // Matrix of the linear Dynamics system float* State; // Vector of State int SamplesNum; // Number of the Samples float** flSamples; // array of the Sample Vectors float** flNewSamples; // temporary array of the Sample Vectors float* flConfidence; // Confidence for each Sample float* flCumulative; // Cumulative confidence float* Temp; // Temporary vector float* RandomSample; // RandomVector to update sample set CvRandState* RandS; // Array of structures to generate random vectors }CvConDensation;

9-20

Image Analysis Reference

Table 10-1 Image Analysis Reference
Group Name Functions Contour Retrieving Functions Description

10
Finds contours in a binary image. Initializes contour scanning process. Finds the next contour on the raster. Replaces the retrieved contour. Finishes scanning process. Calculates convolution of the input image with Laplacian operator. Calculates convolution of the input image with Sobel operator. Implements Canny algorithm for edge detection. Calculates two constraint images for corner detection.

FindContours StartFindContours FindNextContour SubstituteContour EndFindContours

Features Functions

Laplace

Sobel

Canny

PreCornerDetect

10-1

OpenCV Reference Manual

Image Analysis Reference

Table 10-1 Image Analysis Reference (continued)

Group Name Description Calculates eigenvalues and eigenvectors of image blocks for corner detection. Calculates minimal eigenvalues of image blocks for corner detection. Refines corner locations. Determines strong corners on the image. Finds lines in a binary image, SHT algorithm. Finds lines in a binary image, MHT algorithm. Finds line segments in a binary image, PPHT algorithm. Counts non-zero pixels in an image. Summarizes pixel values in an image. Calculates mean value in an image region. Calculates mean and standard deviation in an image region. Finds global minimum and maximum in an image region. Calculates image norm, difference norm or relative difference norm

CornerEigenValsAndVecs

CornerMinEigenVal

FindCornerSubPix GoodFeaturesToTrack HoughLines HoughLinesSDiv HoughLinesP

Image Statistics Functions

CountNonZero SumPixels Mean Mean_StdDev

MinMaxLoc

Norm

10-2

OpenCV Reference Manual

Image Analysis Reference

Table 10-1 Image Analysis Reference (continued)

Group Name Description Calculates all moments up to the third order of the image plane and fills the moment state structure. Retrieves spatial moment from the moment state structure. Retrieves the central moment from the moment state structure. Retrieves the normalized central moment from the moment state structure. Calculates seven Hu moment invariants from the moment state structure. Downsamples an image. Upsamples an image. Implements image segmentation by pyramids. Creates a structuring element. Deletes the structuring element. Erodes the image by using an arbitrary structuring element. Dilates the image by using an arbitrary structuring element. Performs advanced morphological transformations.

Moments

GetSpatialMoment

GetCentralMoment

GetNormalizedCentralMoment

GetHuMoments

Pyramid Functions

PyrDown PyrUp PyrSegmentation

Morphology Functions

CreateStructuringElementEx ReleaseStructuringElement Erode

Dilate

MorphologyEx

10-3

OpenCV Reference Manual

Image Analysis Reference

Table 10-1 Image Analysis Reference (continued)

Group Distance Transform Function Name Description Calculates distance to the closest zero pixel for all non-zero pixels of the source image. Provides an adaptive thresholding binary image. Thresholds the binary image. Makes flood filling of the image connected domain. Creates a histogram. Releases the histogram header and the underlying data. Initializes the histogram header. Queries the value of a 1D histogram bin. Queries the value of a 2D histogram bin Queries the value of a 3D histogram bin Queries the value of an nD histogram bin Returns the pointer to 1D histogram bin. Returns the pointer to 2D histogram bin. Returns the pointer to 3D histogram bin. Returns the pointer to nD histogram bin.

DistTransform

Threshold Functions

AdaptiveThreshold

Threshold
Flood Filling Function Histogram Functions

FloodFill CreateHist ReleaseHist

MakeHistHeaderForArray QueryHistValue_1D QueryHistValue_2D QueryHistValue_3D QueryHistValue_nD GetHistValue_1D GetHistValue_2D GetHistValue_3D GetHistValue_nD

10-4

OpenCV Reference Manual

Image Analysis Reference

Table 10-1 Image Analysis Reference (continued)

Group Name Description Finds minimum and maximum histogram bins. Normalizes a histogram. Thresholds a histogram. Compares two histograms. Makes a copy of a histogram. Sets bounds of histogram bins. Calculates a histogram of an array of single-channel images. Calculates back projection of a histogram. Calculates back projection by comparing histograms of the source image patches with the given histogram. Computes earth mover distance and/or a lower boundary of the distance. Calculates a histogram of contrast for the one-channel image. Represents an element for each single connected components representation in memory. Stores all the types of histograms (1D, 2D, nD).

GetMinMaxHistValue NormalizeHist ThreshHist CompareHist CopyHist SetHistBinRanges CalcHist

CalcBackProject CalcBackProjectPatch

CalcEMD

CalcContrastHist

Data Types Pyramid Data Types

CvConnectedComp

Histogram Data Types

CvHistogram

10-5

OpenCV Reference Manual

Image Analysis Reference

Contour Retrieving Functions

FindContours
Finds contours in binary image.
int cvFindContours(IplImage* img, CvMemStorage* storage, CvSeq** firstContour, int headerSize=sizeof(CvContour), CvContourRetrievalMode mode=CV_RETR_LIST,CvChainApproxMethod method=CV_CHAIN_APPROX_SIMPLE); img

Single channel image of IPL_DEPTH_8U type. Non-zero pixels are treated as 1-pixels. The function modifies the content of the input parameter. Contour storage location. Output parameter. Pointer to the first contour on the highest level. Size of the sequence header; must be equal to or greater than sizeof(CvChain) when the method CV_CHAIN_CODE is used, and equal to or greater than sizeof(CvContour) otherwise. Retrieval mode.

storage firstContour

headerSize

mode

method

CV_RETR_EXTERNAL

retrieves only the extreme outer

contours (list);
CV_RETR_LIST

retrieves all the contours (list);

CV_RETR_CCOMP

retrieves the two-level hierarchy (list of connected components); retrieves the complete hierarchy (tree). contours in the Freeman chain

CV_RETR_TREE

Approximation method.

CV_CHAIN_CODE outputs

code.

10-6

OpenCV Reference Manual

Image Analysis Reference

CV_CHAIN_APPROX_NONE

translates all the points from the chain code into points; compresses horizontal, vertical, and diagonal segments, that is, it leaves only their ending points; are two versions of the Teh-Chin approximation algorithm.

CV_CHAIN_APPROX_SIMPLE

CV_CHAIN_APPROX_TC89_L1, CV_CHAIN_APPROX_TC89_KCOS

Discussion The function FindContours retrieves contours from the binary image and returns the pointer to the first contour. Access to other contours may be gained through the h_next and v_next fields of the returned structure. The function returns total number of retrieved contours.

StartFindContours
Initializes contour scanning process.
CvContourScanner cvStartFindContours(IplImage* img, CvMemStorage* storage, int headerSize, CvContourRetrievalMode mode, CvChainApproxMethod method ); img

Single channel image of IPL_DEPTH_8U type. Non-zero pixels are treated as 1-pixels. The function damages the image. Contour storage location. Must be equal to or greater than sizeof(CvChain) when the method CV_CHAIN_CODE is used, and equal to or greater than sizeof(CvContour) otherwise. Retrieval mode.

storage headerSize

mode

CV_RETR_EXTERNAL

retrieves only the extreme outer

contours (list);

10-7

OpenCV Reference Manual

Image Analysis Reference

method

CV_RETR_LIST

retrieves all the contours (list);

CV_RETR_CCOMP

retrieves the two-level hierarchy (list of connected components); retrieves the complete hierarchy (tree). the output contours in the chain

CV_RETR_TREE

Approximation method.

CV_CHAIN_CODE codes

code;
CV_CHAIN_APPROX_NONE

translates all the points from the chain code into points; horizontal, vertical, and diagonal segments;

CV_CHAIN_APPROX_SIMPLE substitutes ending points for CV_CHAIN_APPROX_TC89_L1, CV_CHAIN_APPROX_TC89_KCOS

are two versions of the Teh-Chin approximation algorithm.

Discussion The function StartFindContours initializes the contour scanner and returns the pointer to it. The structure is internal and no description is provided.

FindNextContour
Finds next contour on raster.
CvSeq* cvFindNextContour( CvContourScanner scanner ); scanner

Contour scanner initialized by the function cvStartFindContours.

Discussion The function FindNextContour returns the next contour or 0, if the image contains no other contours.

10-8

OpenCV Reference Manual

Image Analysis Reference

SubstituteContour
Replaces retrieved contour.
void cvSubstituteContour( CvContourScanner scanner, CvSeq* newContour ); scanner newContour

Contour scanner initialized by the function cvStartFindContours. Substituting contour.

Discussion The function SubstituteContour replaces the retrieved contour, that was returned from the preceding call of the function FindNextContour and stored inside the contour scanner state, with the user-specified contour. The contour is inserted into the resulting structure, list, two-level hierarchy, or tree, depending on the retrieval mode. If the parameter newContour is 0, the retrieved contour is not included into the resulting structure, nor all of its children that might be added to this structure later.

EndFindContours
Finishes scanning process.
CvSeq* cvEndFindContours( CvContourScanner* scanner ); scanner

Pointer to the contour scanner.

Discussion The function EndFindContours finishes the scanning process and returns the pointer to the first contour on the highest level.

10-9

OpenCV Reference Manual

Image Analysis Reference

Features Functions
Fixed Filters Functions
For background on fundamentals of Fixed Filters Functions see Fixed Filters in Image Analysis Chapter.

Laplace
Calculates convolution of input image with Laplacian operator.
void cvLaplace( IplImage* src, IplImage* dst, int apertureSize=3); src dst apertureSize

Input image. Destination image. Size of the Laplacian kernel.

Discussion The function Laplace calculates the convolution of the input image src with the Laplacian kernel of a specified size apertureSize and stores the result in dst.

Sobel
Calculates convolution of input image with Sobel operator.
void cvSobel( IplImage* src, IplImage* dst, int dx, int dy, int apertureSize=3); src dst

Input image. Destination image.

10-10

OpenCV Reference Manual

Image Analysis Reference

dx dy apertureSize

Order of the derivative x. Order of the derivative y. Size of the extended Sobel kernel. The special value CV_SCHARR, equal to -1, corresponds to the Scharr filter 1/16[-3,-10,-3; 0, 0, 0; 3, 10, 3]; may be transposed.

Discussion The function Sobel calculates the convolution of the input image src with a specified Sobel operator kernel and stores the result in dst.

Feature Detection Functions

For background on fundamentals of Feature Detection Functions see Feature Detection in Image Analysis Chapter.

Canny
Implements Canny algorithm for edge detection.
void cvCanny( IplImage* img, IplImage* edges, double lowThresh, double highThresh, int apertureSize=3 ); img edges lowThresh highThresh apertureSize

Input image. Image to store the edges found by the function. Low threshold used for edge searching. High threshold used for edge searching. Size of the Sobel operator to be used in the algorithm.

Discussion The function Canny finds the edges on the input image img and puts them into the output image edges using the Canny algorithm described above.

10-11

OpenCV Reference Manual

Image Analysis Reference

PreCornerDetect
Calculates two constraint images for corner detection.
void cvPreCornerDetect( IplImage* img, IplImage* corners, Int apertureSize ); img corners apertureSize

Input image. Image to store the results. Size of the Sobel operator to be used in the algorithm.

Discussion The function PreCornerDetect finds the corners on the input image img and stores them into the output image corners in accordance with Method 1for corner detection.

CornerEigenValsAndVecs
Calculates eigenvalues and eigenvectors of image blocks for corner detection.
void cvCornerEigenValsAndVecs( IplImage* img, IplImage* eigenvv, int blockSize, int apertureSize=3 ); img eigenvv blockSize

Input image. Image to store the results. Linear size of the square block over which derivatives averaging is done. Derivative operator aperture size in the case of byte source format. In the case of floating-point input format this parameter is the number of the fixed float filter used for differencing.

apertureSize

10-12

OpenCV Reference Manual

Image Analysis Reference

Discussion For every raster pixel the function CornerEigenValsAndVecs takes a block of blockSize blockSize pixels with the top-left corner, or top-bottom corner for bottom-origin images, at the pixel, computes first derivatives Dx and Dy within the block and then computes eigenvalues and eigenvectors of the matrix:

C =

The format of the frame eigenvv is the following: for every pixel of the input image the frame contains 6 float values ( 1, 2, x 1, y 1, x 2, y 2 ).
1, 2
x 1, y 1 x 2, y 2

In case of a singular matrix or if one of the eigenvalues is much less than another, all six values are set to 0. The Sobel operator with aperture width aperureSize is used for differentiation.

CornerMinEigenVal
Calculates minimal eigenvalues of image blocks for corner detection.
void cvCornerMinEigenVal( IplImage* img, IplImage* eigenvv, int blockSize, int apertureSize=3 ); img eigenvv blockSize

DxDy

D x Dy Dy
2

, where summations are performed over the block.

are eigenvalues of the above matrix, not sorted by value. are coordinates of the normalized eigenvector that corresponds to are coordinates of the normalized eigenvector that corresponds to
1 . 2 .

Input image. Image to store the results. Linear size of the square block over which derivatives averaging is done.

10-13

OpenCV Reference Manual

Image Analysis Reference

apertureSize

Derivative operator aperture size in the case of byte source format. In the case of floating-point input format this parameter is the number of the fixed float filter used for differencing.

Discussion For every raster pixel the function CornerMinEigenVal takes a block of blockSize blockSize pixels with the top-left corner, or top-bottom corner for bottom-origin images, at the pixel, computes first derivatives Dx and Dy within the block and then computes eigenvalues and eigenvectors of the matrix:

C =

In case of a singular matrix the minimal eigenvalue is set to 0. The Sobel operator with aperture width aperureSize is used for differentiation.

FindCornerSubPix
Refines corner locations.
void cvFindCornerSubPix( IplImage* img, CvPoint2D32f* corners, int count, CvSize win, CvSize zeroZone, CvTermCriteria criteria ); img corners

count win

zeroZone

DxDy

D x Dy Dy
2

, where summations are made over the block.

Input raster image. Initial coordinates of the input corners and refined coordinates on output. Number of corners. Half sizes of the search window. For example, if win = (5,5), then 5 2 + 1 5 2 + 1 = 11 11 pixel window to be used. Half size of the dead region in the middle of the search zone to avoid possible singularities of the autocorrelation matrix. The value of (-1,-1)indicates that there is no such zone.

10-14

OpenCV Reference Manual

Image Analysis Reference

criteria

Criteria for termination of the iterative process of corner refinement. Iterations may specify a stop when either required precision is achieved or the maximal number of iterations done.

Discussion. The function FindCornerSubPix iterates to find the accurate sub-pixel location of a corner, or radial saddle point, as shown in Figure 10-1.

Figure 10-1 Sub-Pixel Accurate Corner

Sub-pixel accurate corner (radial saddle point) locator is based on the observation that any vector from q to p is orthogonal to the image gradient.

10-15

OpenCV Reference Manual

Image Analysis Reference

The core idea of this algorithm is based on the observation that every vector from the center q to a point p located within a neighborhood of q is orthogonal to the image gradient at p subject to image and measurement noise. Thus:
i = I p ( q pi )
i

where I p i is the image gradient at the one of the points p in a neighborhood of q. The value of q is to be found such that i is minimized. A system of equations may be set up with i s set to zero:
I p I p
i

where the gradients are summed within a neighborhood (search window) of q. Calling the first gradient term G and the second gradient term b gives:
q= G
1

The algorithm sets the center of the neighborhood window at this new center q and then iterates until the center keeps within a set threshold.

GoodFeaturesToTrack
Determines strong corners on image.
void cvGoodFeaturesToTrack( IplImage* image, IplImage* eigImage, IplImage* tempImage, CvPoint2D32f* corners, int* cornerCount, double qualityLevel, double minDistance ); image

Source image with byte, signed byte, or floating-point depth, single channel. Temporary image for minimal eigenvalues for pixels: floating-point, single channel. Another temporary image: floating-point, single channel. Output parameter. Detected corners.

eigImage

tempImage corners

I p I p p i = 0 ,
i i

10-16

OpenCV Reference Manual

Image Analysis Reference

cornerCount qualityLevel

Output parameter. Number of detected corners. Multiplier for the maxmin eigenvalue; specifies minimal accepted quality of image corners. Limit, specifying minimum possible distance between returned corners; Euclidian distance is used.

minDistance

Discussion The function GoodFeaturesToTrack finds corners with big eigenvalues in the image. The function first calculates the minimal eigenvalue for every pixel of the source image and then performs non-maxima suppression (only local maxima in 3x3 neighborhood remain). The next step is rejecting the corners with the minimal eigenvalue less than qualityLevel*<max_of_min_eigen_vals>. Finally, the function ensures that all the corners found are distanced enough from one another by getting two strongest features and checking that the distance between the points is satisfactory. If not, the point is rejected.

Hough Transform Functions

For background on fundamentals of Hough Transform Functions see Hough Transform in Image Analysis Chapter.

HoughLines
Finds lines in binary image, SHT algorithm.
void cvHoughLines ( IplImage* src, double rho, double theta, int threshold, float* lines, int linesNumber); src rho theta threshold

Source image. Radius resolution. Angle resolution. Threshold parameter.

10-17

OpenCV Reference Manual

Image Analysis Reference

lines

Pointer to the array of output lines parameters. The array should have 2*linesNumber elements. Maximum number of lines.

linesNumber

Discussion The function HoughLines implements Standard Hough Transform (SHT) and demonstrates average performance on arbitrary images. The function returns number of detected lines. Every line is characterized by pair ( , ), where is distance from line to point (0,0) and is the angle between the line and horizontal axis.

HoughLinesSDiv
Finds lines in binary image, MHT algorithm.
int cvHoughLinesSDiv ( IplImage* src, double rho, int srn, double theta, stn, int threshold, float* lines, int linesNumber); src rho srn theta stn threshold lines int

Source image. Rough radius resolution. Radius accuracy coefficient, rho/srn is accurate rho resolution. Rough angle resolution. Angle accuracy coefficient, theta/stn is accurate angle resolution. Threshold parameter. Pointer to array of the detected lines parameters. The array should have 2*linesNumber elements. Maximum number of lines.

linesNumber

Discussion
The function HoughLinesSDiv implements coarse-to-fine variant of SHT and is significantly faster than the latter on images without noise and with a small number of lines. The output of the function has the same format as the output of the function HoughLines.

10-18

OpenCV Reference Manual

Image Analysis Reference

HoughLinesP
Finds line segments in binary image, PPHT algorithm.
int cvHoughLinesP( IplImage* src, double rho, double theta, int threshold, int lineLength, int lineGap, int* lines, int linesNumber ); src rho theta threshold lineLength lineGap lines

Source image. Rough radius resolution. Rough angle resolution. Threshold parameter. Minimum accepted line length. Maximum length of accepted line gap. Pointer to array of the detected line segments' ending coordinates. The array should have linesNumber*4 elements. Maximum number of line segments.

linesNumber

Discussion
The function HoughLinesP implements Progressive Probabilistic Standard Hough Transform. It retrieves no more than linesNumber line segments; each of those must be not shorter than lineLength pixels. The method is significantly faster than SHT on noisy images, containing several long lines. The function returns number of detected segments. Every line segment is characterized by the coordinates of its ends(x1,y1,x2,y2).

10-19

OpenCV Reference Manual

Image Analysis Reference

Image Statistics Functions

CountNonZero
Counts non-zero pixels in image.
int cvCountNonZero (IplImage* image ); image

Pointer to the source image.

Discussion The function CountNonZero returns the number of non-zero pixels in the whole image or selected image ROI.

SumPixels
Summarizes pixel values in image.
double cvSumPixels( IplImage* image ); image

Pointer to the source image.

Discussion The function SumPixels returns sum of pixel values in the whole image or selected image ROI.

10-20

OpenCV Reference Manual

Image Analysis Reference

Mean
Calculates mean value in image region.
double cvMean( IplImage* image, IplImage* mask=0 ); image mask

Pointer to the source image. Mask image.

Discussion The function Mean calculates the mean of pixel values in the whole image, selected ROI or, if mask is not NULL,in an image region of arbitrary shape.

Mean_StdDev
Calculates mean and standard deviation in image region.
void cvMean_StdDev( IplImage* image, double* mean, double* stddev, IplImage* mask=0 ); image mean stddev mask

Pointer to the source image. Pointer to returned mean. Pointer to returned standard deviation. Pointer to the single-channel mask image.

Discussion The function Mean_StdDev calculates mean and standard deviation of pixel values in the whole image, selected ROI or, if mask is not NULL,in an image region of arbitrary shape. If the image has more than one channel, the COI must be selected.

10-21

OpenCV Reference Manual

Image Analysis Reference

MinMaxLoc
Finds global minimum and maximum in image region.
void cvMinMaxLoc( IplImage* image, double* minVal, double* maxVal, CvPoint* minLoc, CvPoint* maxLoc, IplImage* mask=0 ); image minVal maxVal minLoc maxLoc mask

Pointer to the source image. Pointer to returned minimum value. Pointer to returned maximum value. Pointer to returned minimum location. Pointer to returned maximum location. Pointer to the single-channel mask image.

Discussion The function MinMaxLoc finds minimum and maximum pixel values and their positions. The extremums are searched over the whole image, selected ROI or, if mask is not NULL,in an image region of arbitrary shape. If the image has more than one channel, the COI must be selected.

Norm
Calculates image norm, difference norm or relative difference norm.
double cvNorm( IplImage* imgA, IplImage* imgB, int normType, IplImage* mask=0 ); imgA imgA normType

Pointer to the first source image. Pointer to the second source image if any, NULL otherwise. Type of norm.

10-22

OpenCV Reference Manual

Image Analysis Reference

mask

Pointer to the single-channel mask image.

Discussion The function Norm calculates images norms defined below. If imgB = NULL, the following three norm types of image A are calculated:
NormType = CV_C: A
C

= max ( A ij ) ,
Nx Ny

i = 1j = 1

imgB NULL,

the difference or relative difference norms are calculated:

AB
C

NormType = CV_C:

= max ( A i B i ) ,
Nx Ny

i = 1j = 1

NormType = CV_RELATIVEC:

NormType = CV_RELATIVEL1

j = i = 1N = 1N -------------------------------------------------- , x y

i = 1j = 1

10-23

NormType = CV_L2: A B

NormType = CV_L1: A B

NormType = CV_L2: A

Nx Nx

NormType = CV_L1:

A ij

A ij .

A ij B ij

( A ij B ij ) ,
max ( A ij B ij ) = --------------------------------------------- , max ( B ij )
Nx Ny

A ij B ij

Bij

OpenCV Reference Manual

Image Analysis Reference

( A ij B ij )
Ny

NormType = CV_RELATIVEL2:

i = 1j = 1 = ------------------------------------------------------------ . Nx

i = 1j = 1

The function Norm returns the calculated norm.

Moments
Calculates all moments up to third order of image plane and fills moment state structure.
void cvMoments( IplImage* image, CvMoments* moments, int isBinary=0 ); image moments isBinary

Pointer to the image or to top-left corner of its ROI. Pointer to returned moment state structure. If the flag is non-zero, all the zero pixel values are treated as zeroes, all the others are treated as ones.

Discussion The function Moments calculates moments up to the third order and writes the result to the moment state structure. This structure is used then to retrieve a certain spatial, central, or normalized moment or to calculate Hu moments.

10-24

( B ij )

OpenCV Reference Manual

Image Analysis Reference

GetSpatialMoment
Retrieves spatial moment from moment state structure.
double cvGetSpatialMoment( CvMoments* moments, int x_order, int y_order ); moments x_order y_order

Pointer to the moment state structure. Order x of required moment. Order y of required moment (0<= x_order, y_order; x_order + y_order <= 3).

Discussion The function GetSpatialMoment retrieves the spatial moment, which is defined as:
x, y x_order y_order

I ( x, y )

is the intensity of the pixel (x,y).

GetCentralMoment
Retrieves central moment from moment state structure.
double cvGetCentralMoment( CvMoments* moments, int x_order, int y_order ); moments x_order y_order

Pointer to the moment state structure. Order x of required moment. Order y of required moment (0<= x_order, y_order; x_order + y_order <= 3).

M x_order, y_order =

I ( x, y ) x

, where

10-25

OpenCV Reference Manual

Image Analysis Reference

Discussion The function GetCentralMoment retrieves the central moment, which is defined as:
x, y x_order y_order

is the intensity of pixel (x,y), x is the coordinate x of the mass center, y is the coordinate y of the mass center:
M 1, 0 M 0, 1 x = ---------- , y = ---------- . M 0, 0 M 0, 0

I ( x, y )

GetNormalizedCentralMoment
Retrieves normalized central moment from moment state structure.
double cvGetNormalizedCentralMoment(CvMoments* moments, int x_order, int y_order); moments x_order y_order

Pointer to the moment state structure. Order x of required moment. Order y of required moment (0<= x_order, y_order; x_order + y_order <= 3).

Discussion The function GetNormalizedCentralMoment retrieves the normalized central moment, which is defined as:
x_order, y_order x_order, y_order = ----------------------------------------------------------------------) . ( ( x_order + y_order ) 2 + 1 M 0, 0

x_order, y_order =

I ( x, y ) ( x x )

(y y)

, where

10-26

OpenCV Reference Manual

Image Analysis Reference

GetHuMoments
Calculates seven moment invariants from moment state structure.
void cvGetHuMoments( CvMoments* moments, CvHuMoments* HuMoments ); moments HuMoments

Pointer to the moment state structure. Pointer to Hu moments structure.

Discussion The function GetHuMoments calculates seven Hu invariants using the following formulas:
h 1 = 20 + 02 , h 2 = ( 20 02 ) + 4 11 , h 3 = ( 30 3 12 ) + ( 3 21 03 ) h 4 = ( 30 + 12 ) + ( 21 + 03 )
2 2 2 2 2 2

,
2 2

h 5 = ( 30 3 12 ) ( 30 + 12 ) [ ( 30 + 12 ) 3 ( 21 + 03 ) ]

Input srcImage and output dstImage should have the identical IPL_DEPTH_8U depth and identical number of channels (1 or 3).

Morphology Functions

CreateStructuringElementEx
Creates structuring element.
IplConvKernel* cvCreateStructuringElementEx(int nCols, int nRows, int anchorX, int anchorY, CvElementShape shape, int* values ); nCols nRows anchorX anchorY shape

Number of columns in the structuring element. Number of rows in the structuring element. Relative horizontal offset of the anchor point. Relative vertical offset of the anchor point. Shape of the structuring element; may have the following values:

CV_SHAPE_RECT,

a rectangular element; a cross-shaped element; an elliptic element;

CV_SHAPE_CROSS,

CV_SHAPE_ELLIPSE,

CV_SHAPE_CUSTOM, a user-defined element. In this case the parameter values specifies the mask, that is, which neighbors of the pixel must be considered.

values

Pointer to the structuring element data, a plane array, representing row-by-row scanning of the element matrix. Non-zero values indicate points that belong to the element. If the pointer is NULL, then all values are considered non-zero, that is, the element is of a rectangular shape. This parameter is considered only if the shape is CV_SHAPE_CUSTOM.

10-30

OpenCV Reference Manual

Image Analysis Reference

Discussion The function CreateStructuringElementEx allocates and fills the structure IplConvKernel, which can be used as a structuring element in the morphological operations.

ReleaseStructuringElement
Deletes structuring element.
void cvReleaseStructuringElement(IplConvKernel** ppElement); ppElement

Pointer to the deleted structuring element.

Discussion The function ReleaseStructuringElement releases the structure IplConvKernel that is no longer needed. If *ppElement is NULL, the function has no effect. The function returns created structuring element.

Erode
Erodes image by using arbitrary structuring element.
void cvErode( IplImage* src, IplImage* dst, IplConvKernel* B, int iterations); src dst B

Source image. Destination image. Structuring element used for erosion. If NULL, a 3x3 rectangular structuring element is used. Number of times erosion is applied.

iterations

10-31

OpenCV Reference Manual

Image Analysis Reference

Discussion The function Erode erodes the source image. The function takes the pointer to the structuring element, consisting of zeros and minus ones; the minus ones determine neighbors of each pixel from which the minimum is taken and put to the corresponding destination pixel. The function supports the in-place mode when the source and destination pointers are the same. Erosion can be applied several times (iterations parameter). Erosion on a color image means independent transformation of all the channels.

Dilate
Dilates image by using arbitrary structuring element.
void cvDilate( IplImage* pSrc, IplImage* pDst, IplConvKernel* B, int iterations); pSrc pDst B

Source image. Destination image. Structuring element used for dilation. If NULL, a 3x3 rectangular structuring element is used. Number of times dilation is applied.

iterations

Discussion The function Dilate performs dilation of the source image. It takes pointer to the structuring element that consists of zeros and minus ones; the minus ones determine neighbors of each pixel from which the maximum is taken and put to the corresponding destination pixel. The function supports in-place mode. Dilation can be applied several times (iterations parameter). Dilation of a color image means independent transformation of all the channels.

10-32

OpenCV Reference Manual

Image Analysis Reference

MorphologyEx
Performs advanced morphological transformations.
void cvMorphologyEx( IplImage* src, IplImage* dst, IplImage* temp, IplConvKernel* B, CvMorphOp op, int iterations ); src dst temp B op

Source image. Destination image. Temporary image, required in some cases. Structuring element. Type of morphological operation:

iterations

CV_MOP_OPEN,

opening; closing; morphological gradient; black hat. top hat;

CV_MOP_CLOSE,

CV_MOP_GRADIENT, CV_MOP_TOPHAT,

CV_MOP_BLACKHAT,

(See Morphology for description of these operations). Number of times erosion and dilation are applied during the complex operation.

Discussion The function MorphologyEx performs advanced morphological transformations. The function uses Erode and Dilate to perform more complex operations. The parameter temp must be non-NULL and point to the image of the same size and format as src and dst when op is CV_MOP_GRADIENT, or when op is CV_MOP_TOPHAT or op is CV_MOP_BLACKHAT and src is equal to dst (in-place operation).

10-33

OpenCV Reference Manual

Image Analysis Reference

Distance Transform Function

DistTransform
Calculates distance to closest zero pixel for all non-zero pixels of source image.
void cvDistTransform ( IplImage* src, IplImage* dst, CvDisType disType, CvDisMaskType maskType, float* mask); src dst disType

Source image. Output image with calculated distances. Type of distance; can be CV_DIST_L1, CV_DIST_L2, CV_DIST_C or CV_DIST_USER. Size of distance transform mask; can be CV_DIST_MASK_3x3 or CV_DIST_MASK_5x5. Pointer to the user-defined mask used with the distance type CV_DIST_USER.

maskType

mask

10-34

OpenCV Reference Manual

Image Analysis Reference

Discussion The function DistTransform approximates the actual distance from the closest zero pixel with a sum of fixed distance values: two for 3x3 mask and three for 5x5 mask. Figure 10-2 shows the result of the distance transform of a 7x7 image with a zero central pixel.
Figure 10-2 3x3 Mask

4.5 4 3.5 3 3.5 4 4.5

4 3 2.5 2 2.5 3 4

3.5 2.5 1.5 1 1.5 2.5 3.5

3 2 1 0 1 2 3

3.5 2.5 1.5 1 1.5 2.5 3.5

4 3 2.5 2 2.5 3 4

4.5 4 3.5 3 3.5 4 4.5

This example corresponds to a 3x3 mask; in case of user-defined distance type the user sets the distance between two pixels, that share the edge, and the distance between the pixels, that share the corner only. For this case the values are 1 and 1.5 correspondingly. Figure 10-3 shows the distance transform for the same image, but for a 5x5 mask. For the 5x5 mask the user sets an additional distance that is the distance between pixels corresponding to the chess knight move. In this example the additional distance is equal to 2. For CV_DIST_L1, CV_DIST_L2, and CV_DIST_C the optimal precalculated distance values are used.

10-35

OpenCV Reference Manual

Image Analysis Reference

Figure 10-3 5x5 Mask

4.5 3.5 3 3 3 3.5 4

3.5 3 2 2 2 3 3.5

3 2 1.5 1 1.5 2 3

3 2 1 0 1 2 3

3 2 1.5 1 1.5 2 3

3.5 3 2 2 2 3 3.5

4 3.5 3 3 3 3.5 4

Threshold Functions

AdaptiveThreshold
Provides adaptive thresholding binary image.
void cvAdaptiveThreshold( IplImage* src, IplImage* dst, double max, CvAdaptiveThreshMethod method, CvThreshType type, double* parameters); src dst

Source image. Destination image.

10-36

OpenCV Reference Manual

Image Analysis Reference

max

Max parameter used with the types CV_THRESH_BINARY and CV_THRESH_BINARY_INV only. Method for the adaptive threshold definition; now CV_STDDEF_ADAPTIVE_THRESH only. Thresholding type; must be one of

method

type

parameters

CV_THRESH_BINARY, val = ( val > Thresh ?MAX:0 ) ; CV_THRESH_BINARY_INV, val = ( val > Thresh ?0:MAX ) ; CV_THRESH_TOZERO, val = ( val > Thresh ? val :0 ) ; CV_THRESH_TOZERO_INV, val = ( val > Thresh ?0: val ) .

Pointer to the list of method-specific input parameters. For the method CV_STDDEF_ADAPTIVE_THRESH the value parameters[0] is the size of the neighborhood: 1-(3x3), 2-(5x5), or 3-(7x7), and parameters[1] is the value of the minimum variance.

Discussion The function AdaptiveThreshold calculates the adaptive threshold for every input image pixel and segments image. The algorithm is as follows. Let { f ij }, 1 i l, 1 j J be the input image. For every pixel variance v ij are calculated as follows:
p p p p

i, j

the mean

m ij

and

s = p t = p

where

is the neighborhood.

Local threshold for pixel i, j is t ij = m ij + v ij for v ij > v min , and t ij = t ij 1 for v ij v min , where v min is the minimum variance value. If j = 1 , then t ij = t i 1, j , t 11 = t i j , where v i j > v min and vij v min for ( i < i 0 ) ( ( i = i 0 ) ( j < j 0 ) ) . 0 0 0 0 Output segmented image is calculated as in the function Threshold.

10-37

m ij = 1 2 p

f i + s, j + t , v ij = 1 2 p

f i + s, j + t m ij

OpenCV Reference Manual

Image Analysis Reference

Threshold
Thresholds binary image.
void cvThreshold( IplImage* src, IplImage* dst, float thresh, float maxvalue, CvThreshType type); src dst thresh maxvalue

Source image. Destination image; can be the same as the parameter src. Threshold parameter. Maximum value; parameter, used with threshold types CV_THRESH_BINARY, CV_THRESH_BINARY_INV, and CV_THRESH_TRUNC. Thresholding type; must be one of

type

Discussion

CV_THRESH_BINARY, val = ( val > thresh maxvalue :0 ) ; CV_THRESH_BINARY_INV, val = ( val > thresh 0 :maxvalue ) ; CV_THRESH_TRUNC, val = ( val > thresh?thresh:maxvalue ) ; CV_THRESH_TOZERO, val = ( val > thresh val :0 ) ; CV_THRESH_TOZERO_INV, val = ( val > thresh 0: val ) .

The function Threshold applies fixed-level thresholding to grayscale image. The result is either a grayscale image or a bi-level image. The former variant is typically used to remove noise from the image, while the latter one is used to represent a grayscale image as composition of connected components and after that build contours on the components via the function FindContours. Figure 10-4 illustrates meanings of different threshold types:

10-38

OpenCV Reference Manual

Image Analysis Reference

Figure 10-4 Meanings of Threshold Types

10-39

OpenCV Reference Manual

10-47

OpenCV Reference Manual

Image Analysis Reference

factor

Normalization factor.

Discussion The function NormalizeHist normalizes the histogram, such that the sum of histogram bins becomes equal to factor.

ThreshHist
Thresholds histogram.
void cvThreshHist( CvHistogram* hist, float thresh); hist thresh

Pointer to the histogram. Threshold level.

Discussion The function ThreshHist clears histogram bins that are below the specified level.

CompareHist
Compares two histograms.
double cvCompareHist( CvHistogram* hist1, CvHistogram* hist2, CvCompareMethod method ); hist1 hist2 method

First histogram. Second histogram. Comparison method; may be any of those listed below:

CV_COMP_CORREL; CV_COMP_CHISQR; CV_COMP_INTERSECT.

10-48

OpenCV Reference Manual

Image Analysis Reference

Discussion The function CompareHist compares two histograms using specified method.

i CV_COMP_CORREL result = ---------------------------------- , 2 2

The function returns the comparison result.

CopyHist
Copies histogram.
void cvCopyHist( CvHistogram* src, CvHistogram** dst ); src dst

Source histogram. Pointer to destination histogram.

Discussion The function CopyHist makes a copy of the histogram. If the second histogram pointer *dst is null, it is allocated and the pointer is stored at *dst. Otherwise, both histograms must have equal types and sizes, and the function simply copies the source histogram bins values to destination histogram.

CV_COMP_INTERSECT result =

CV_COMP_CHISQR result =

qi
i i i

qi v i

vi
2

( qi v i ) ------------------------- , q i + vi
min ( q i, v i ) .

10-49

OpenCV Reference Manual

Image Analysis Reference

SetHistBinRanges
Sets bounds of histogram bins.
void cvSetHistBinRanges( CvHistogram* hist, float** ranges, int uniform=1 ); hist ranges uniform

CalcEMD
Computes earth mover distance.
void cvCalcEMD(float* signature1, int size1, float* signature2, int size2, int dims, CvDisType distType, float *distFunc (float* f1, float* f2, void* userParam), float* emd, float* lowerBound, void* userParam); signature1 size1 signature2 size2 dims distType

First signature, array of size1 * (dims + 1) elements. Number of elements in the first compared signature. Second signature, array of size2 * (dims + 1) elements. Number of elements in the second compared signature. Number of dimensions in feature space. Metrics used; CV_DIST_L1, CV_DIST_L2, and CV_DIST_C stand for one of the standard metrics. CV_DIST_USER means that a user-defined function is used as the metric. The function takes two coordinate vectors and user parameter and returns the distance between two vectors. Pointer to the user-defined ground distance function if distType is CV_DIST_USER. Pointer to the calculated emd distance. Pointer to the calculated lower boundary. Pointer to optional data that is passed into the distance function.

distFunc

emd lowerBound userParam

Discussion The function CalcEMD computes earth mover distance and/or a lower boundary of the distance. The lower boundary can be calculated only if dims > 0, and it has sense only if the metric used satisfies all metric axioms. The lower boundary is calculated very fast and can be used to determine roughly whether the two signatures are far enough so that they cannot relate to the same object. If the parameter dims is equal to 0, then signature1 and signature2 are considered simple 1D histograms. Otherwise, both signatures must look as follows:

10-54

OpenCV Reference Manual

Image Analysis Reference

(weight_i0, x0_i0, x1_i0, ..., x(dims-1)_i0,

weight_i1, x0_i1, x1_i1, ..., x(dims-1)_i1, weight_(size1-1), x0_(size1-1), x1_(size1-1, ..., x(dims-1)_(size1-1)),

where weight_ik is the weight of ik cluster, while x0_ik,..., x(dims-1)_ik are coordinates of the cluster ik. If the parameter lower_bound is equal to 0, only emd is calculated. If the calculated lower boundary is greater than or equal to the value stored at this pointer, then the true emd is not calculated, but is set to that lower_bound.

CalcContrastHist
Calculates histogram of contrast.
void cvCalcContrastHist( IplImage **src, CvHistogram* hist, int dontClear, IplImage* mask); src
hist

Pointer to the source images, (now only src[0] is used). Destination histogram. Clear flag. Mask image.

dontClear mask

Discussion The function CalcContrastHist calculates a histogram of contrast for the one-channel image. If dont_clear parameter is 0 then the histogram is cleared before calculation, otherwise it is simply updated. The algorithm works as follows. Let S be a set of pairs (x1, x2) of neighbor pixels in the image f(x)and S ( t ) = { (x 1,x 2) S, f ( x 1 ) t < f ( x 2 ) f ( x 2 ) t < f ( x 1 ) } . Lets denote
{Gt }

as the destination histogram,

10-55

OpenCV Reference Manual

Image Analysis Reference

Et Nt Then
Nt = S ( t ) , Et =

as the summary contrast corresponding to the threshold t, as the counter of the edges detected by the threshold t.
C ( x 1 ,x 2 ,t ) ,

(x1 ,x 2) S ( t )

where

C ( x1 ,x2 ,t ) = min { f ( x 1 ) t , f ( x 2 ) t } and

as
Gt = E t N t, N t 0,

0, N t = 0.

If pointer to the mask is NULL, the histogram is calculated for the all image pixels. Otherwise only those pixels are considered that have non-zero value in the mask in the corresponding position.

Pyramid Data Types

The pyramid functions use the data structure IplImage for image representation and the data structure CvSeq for the sequence of the connected components representation. Every element of this sequence is the data structure CvConnectedComp for the single connected component representation in memory. The C language definition for the CvConnectedComp structure is given below.
Example 10-1 CvConnectedComp
typedef struct CvConnectedComp { double area; /* area of the segmented component */ float value; /* gray scale value of the segmented component */ CvRect rect; /* ROI of the segmented component */ } CvConnectedComp;

the resulting histogram is calculated

10-56

OpenCV Reference Manual

Image Analysis Reference

Histogram Data Types

Example 10-2 CvHistogram
typedef struct CvHistogram { int header_size; /* header's size */ CvHistType type; /* type of histogram */ int flags; /* histograms flags */ int c_dims; /* histograms dimension */ int dims[CV_HIST_MAX_DIM]; /* every dimension size */ int mdims[CV_HIST_MAX_DIM]; /* coefficients for fast access to element */ /* &m[a,b,c] = m + a*mdims[0] + b*mdims[1] + c*mdims[2] */ float* thresh[CV_HIST_MAX_DIM]; /* bin boundaries arrays for every dimension */ float* array; /* all the histogram data, expanded into the single row */ struct CvNode* root; /* tree histogram data */ CvSet* set; /* pointer to memory storage (for tree data) */ int* chdims[CV_HIST_MAX_DIM]; /* cache data for fast calculating */ } CvHistogram;

10-57

OpenCV Reference Manual

Image Analysis Reference

10-58

Structural Analysis Reference

Table 11-1 Structural Analysis Functions
Group Contour Processing Functions Name Functions Description

11
Approximates Freeman chain(s) with a polygonal curve. Initializes the chain reader. Returns the current chain point and moves to the next point. Approximates one or more contours with desired precision. Draws contour outlines in the image. Calculates the bounding box of the contour. Calculates unnormalized spatial and central moments of the contour up to order 3. Calculates the region area within the contour or contour section.

ApproxChains

StartReadChainPoints ReadChainPoint

ApproxPoly

DrawContours ContourBoundingRect ContoursMoments

ContourArea

11-1

OpenCV Reference Manual

Structural Analysis Reference

Table 11-1 Structural Analysis Functions (continued)

Group Name Description Calculates one of the three similarity measures between two contours. Creates binary tree representation for the input contour and returns the pointer to its root. Restores the contour from its binary tree representation. Calculates the value of the matching measure for two contour trees. Fits an ellipse to a set of 2D points. Fits a 2D line to a set of points on the plane. Fits a 3D line to a set of points on the plane. Provides a general way of projecting a set of 3D points to a 2D plane. Finds the convex hull of a set of points. Finds the convex hull of a set of points returning

MatchContours

CreateContourTree

ContourFromContourTree

MatchContourTrees

Geometry Functions

FitEllipse FitLine2D FitLine3D Project3D

ConvexHull ContourConvexHull

cvSeq. ConvexHullApprox
Finds approximate convex hull of a set of points. Finds approximate convex hull of a set of points returning

ContourConvexHullApprox

cvSeq.

11-2

OpenCV Reference Manual

Structural Analysis Reference

Table 11-1 Structural Analysis Functions (continued)

Group Name Description Tests whether the input is a contour convex or not. Finds all convexity defects of the input contour. Finds a circumscribed rectangle of the minimal area for a given convex contour. Calculates a pair-wise geometrical histogram for the contour. Finds the minimal enclosing circle for the planar point set. Represents the contour binary tree in memory. Represents the convexity defect.

CheckContourConvexity

ConvexityDefects

MinAreaRect

CalcPGH

MinEnclosingCircle

Data Types Contour Processing Data Types Geometry Data Types

CvContourTree CvConvexityDefect

Contour Processing Functions

ApproxChains
Approximates Freeman chain(s) with polygonal curve.
CvSeq* cvApproxChains( CvSeq* srcSeq, CvMemStorage* storage, CvChainApproxMethod method=CV_CHAIN_APPROX_SIMPLE, float parameter=0,int minimalPerimeter=0, int recursive=0);

11-3

OpenCV Reference Manual

Structural Analysis Reference

srcSeq storage method

Pointer to the chain that can refer to other chains. Storage location for the resulting polylines. Approximation method (see the description of the function FindContours). Method parameter (not used now). Approximates only those contours whose perimeters are not less than minimalPerimeter. Other chains are removed from the resulting structure. If not 0, the function approximates all chains that access can be obtained to from srcSeq by h_next or v_next links. If 0, the single chain is approximated.

parameter minimalPerimeter

recursive

Discussion This is a stand-alone approximation routine. The function ApproxChains works exactly in the same way as the functions FindContours / FindNextContour with the corresponding approximation flag. The function returns pointer to the first resultant contour. Other contours, if any, can be accessed via v_next or h_next fields of the returned structure.

StartReadChainPoints
Initializes chain reader.
void cvStartReadChainPoints( CvChain* chain, CvChainPtReader* reader ); chain reader

Pointer to chain. Chain reader state.

Discussion The function StartReadChainPoints initializes a special reader (see Dynamic Data Structures for more information on sets and sequences).

11-4

OpenCV Reference Manual

Structural Analysis Reference

ReadChainPoint
Gets next chain point.
CvPoint cvReadChainPoint( CvChainPtReader* reader); reader

Chain reader state.

Pointer to the input contour header. Pointer to the output structure of contour moments; must be allocated by the caller.

Discussion The function ContoursMoments calculates unnormalized spatial and central moments of the contour up to order 3.

ContourArea
Calculates region area inside contour or contour section.
double cvContourSecArea(CvSeq* contour, CvSlice slice=CV_WHOLE_SEQ(seq)); contour slice

Pointer to the input contour header. Starting and ending points of the contour section of interest.

Discussion The function ContourSecArea calculates the region area within the contour consisting of n points p i = ( x i , y i ) , 0 i n , p0 = pn , as a spatial moment:
n

i=1

00 = 1 2

xi 1 yi xiyi 1 .

11-8

OpenCV Reference Manual

Structural Analysis Reference

If a part of the contour is selected and the chord, connecting ending points, intersects the contour in several places, then the sum of all subsection areas is calculated. If the input contour has points of self-intersection, the region area within the contour may be calculated incorrectly.

MatchContours
Matches two contours.
double cvMatchContours (CvSeq *contour1, CvSeq* contour2,int method, long parameter=0 ); contour1 contour2 parameter method

Pointer to the first input contour header. Pointer to the second input contour header. Method-specific parameter, currently ignored. Method for the similarity measure calculation; must be any of

Discussion

CV_CONTOURS_MATCH_I1; CV_CONTOURS_MATCH_I2; CV_CONTOURS_MATCH_I3.

The function MatchContours calculates one of the three similarity measures between two contours. Let two closed contours A and B have n and m points respectively: Normalized central moments of a contour may be denoted as pq , 0 p + q 3 . M. Hu has shown that a set of the next seven features derived from the second and third moments of contours is an invariant to translation, rotation, and scale change [Hu62].
h 1 = 20 + 02 A = { ( xi , y i ) , 1 i n } B = { ( u i , v i ) , 1 i m } .

,
2 2

h 2 = ( 20 02 ) + 4 11

11-9

OpenCV Reference Manual h 3 = ( 30 3 12 ) + ( 3 21 03 ) h 4 = ( 30 + 12 ) + ( 21 + 03 )

2 2 2 2

Structural Analysis Reference

,
2 2

h 5 = ( 30 3 12 ) ( 30 + 12 ) [ ( 30 + 12 ) 3 ( 21 + 03 ) ]

+ ( 3 21 03 ) ( 21 + 03 ) [ 3 ( 30 + 12 ) ( 21 + 03 ) ] ,
h 6 = ( 20 02 ) [ ( 30 + 12 ) ( 21 + 03 ) ] + 4 11 ( 30 + 12 ) ( 21 + 03 ) h 7 = ( 3 21 03 ) ( 30 + 12 ) [ ( 30 + 12 ) 3 ( 21 + 03 ) ]
2 2 2 2

2 2 + ( 30 3 12 ) ( 21 + 03 ) 3 ( 30 + 12 ) ( 21 + 03 ) ] .

From these seven invariant features the three similarity measures I1, I2, and I3 may be calculated: 7
i=1 7

i=1

I 3 ( A , B ) = max ( m i m i ) m i
i

where

m i = sgn ( h i ) log h i
10

CreateContourTree
Creates binary tree representation for input contour.
CvContourTree* cvCreateContourTree(CvSeq *contour, CvMemStorage* storage, double threshold); contour storage threshold

I2( A, B) =

I1( A, B) =

1 mi + 1 mi

mi + mi

,
A

,
, m i = sgn ( h i ) log h i
10 B B B

Pointer to the input contour header. Pointer to the storage block. Value of the threshold.

11-10

OpenCV Reference Manual

Structural Analysis Reference

Discussion The function CreateContourTree creates binary tree representation for the input contour contour and returns the pointer to its root. If the parameter threshold is less than or equal to 0, the function creates full binary tree representation. If the threshold is more than 0, the function creates representation with the precision threshold: if the vertices with the interceptive area of its base line are less than threshold, the tree should not be built any further. The function returns created tree.

ContourFromContourTree
Restores contour from binary tree representation.
CvSeq* cvContourFromContourTree (CvContourTree *tree, CvMemStorage* storage, CvTermCriteria criteria); tree storage criteria

Pointer to the input tree. Pointer to the storage block. Criteria for the definition of the threshold value for contour reconstruction (level of precision).

Discussion The function ContourFromContourTree restores the contour from its binary tree representation. The parameter criterion defines the threshold, that is, level of precision for the contour restoring. If criterion.type = CV_TERMCRIT_ITER, the function restores criterion. maxIter tree levels. If criterion.type = CV_TERMCRIT_EPS, the function restores the contour as long as tri_area > criterion. epsilon *contour_area, where contour_area is the magnitude of the contour area and tri_area is the magnitude of the current triangle area. If criterion.type = CV_TERMCRIT_EPS + CV_TERMCRIT_ITER, the function restores the contour as long as one of these conditions is true. The function returns reconstructed contour.

11-11

OpenCV Reference Manual

Structural Analysis Reference

MatchContourTrees
Compares two binary tree representations.
double cvMatchContourTrees (CvContourTree *tree1, CvContourTree *tree2, CvTreeMatchMethod method, double threshold); tree1 tree2 method

Pointer to the first input tree. Pointer to the second input tree. Method for calculation of the similarity measure; now must be only CV_CONTOUR_TREES_MATCH_I1. Value of the compared threshold.

threshold

Discussion The function MatchContourTrees calculates the value of the matching measure for two contour trees. The similarity measure is calculated level by level from the binary tree roots. If the total calculating value of the similarity for levels from 0 to the specified one is more than the parameter threshold, the function stops calculations and value of the total similarity measure is returned as result. If the total calculating value of the similarity for levels from 0 to the specified one is less than or equal to threshold, the function continues calculation on the next tree level and returns the value of the total similarity measure for the binary trees.

Geometry Functions

FitEllipse
Fits ellipse to set of 2D points.
void cvFitEllipse( CvPoint* points, int n, CvBox2D32f* box ); points

Pointer to the set of 2D points.

11-12

OpenCV Reference Manual

Structural Analysis Reference

n box

Number of points; must be more than or equal to 6. Pointer to the structure for representation of the output ellipse.

Discussion The function FitEllipse fills the output structure in the following way:
boxcenter boxsize boxangle

Point of the center of the ellipse; Sizes of two ellipse axes; Angle between the horizontal axis and the ellipse axis with the length of box->size.width.

The output ellipse has the property of boxsize.width > boxsize.height.

FitLine2D
Fits 2D line to set of points on the plane.
void cvFitLine2D ( CvPoint2D32f* points, int count, CvDisType disType, void* param, float reps, float aeps, float* line); points count disType param

Array of 2D points. Number of points. Type of the distance used to fit the data to a line. Pointer to a user-defined function that calculates the weights for the type CV_DIST_USER, or the pointer to a float user-defined metric parameter c for the Fair and Welsch distance types. Used for iteration stop criteria. If zero, the default value of 0.01 is used. Pointer to the array of four floats. When the function exits, the first two elements contain the direction vector of the line normalized to 1, the other two contain coordinates of a point that belongs to the line.

reps, aeps line

11-13

OpenCV Reference Manual

Structural Analysis Reference

Discussion The function FitLine2D fits a 2D line to a set of points on the plane. Possible distance type values are listed below.
CV_DIST_L2 CV_DIST_L1 CV_DIST_L12 CV_DIST_FAIR CV_DIST_WELSCH CV_DIST_USER c =1.3998.

Standard least squares ( x )

= x

c x ( x ) = ----- 1 exp -c 2

,c = 2.9846.

Uses a user-defined function to calculate the weight. The parameter param should point to the function.
= 0,

The line equation is [ V ( r r 0 ) ] = ( l ine [ 3 ] , line [ 4 ] , line [ 5 . In this algorithm

W ( d ( r i ) ) ri r0

where =

( l ine [ 0 ] , line [ 1 ] , line [ 2 V = 1 ,

and

is the mean of the input vectors with weights, that is,

i r 0 = -------------------------------------. W ( d ( ri ) ) i

The parameters reps and aeps are iteration thresholds. If the distance of the type CV_DIST_C between two values of r 0 calculated from two iterations is less than the value of the parameter reps and the angle in radians between two vectors V is less than the parameter aeps, then the iteration is stopped. The specification for the user-defined weight function is
void userWeight ( float* dist, int count, float* w ); dist count w

The function should fill the weights array with values of weights calculated from the 1 d distance valuesw [ i ] = f ( d [ i ] ) The function f ( x ) = --- ------- has to be monotone decreasing. .
x dx

Pointer to the array of distance values. Number of elements. Pointer to the output array of weights.

11-14

OpenCV Reference Manual

Structural Analysis Reference

FitLine3D
Fits 3D line to set of points in 3D space.
void cvFitLine3D ( CvPoint3D32f* points, int count, CvDisType disType, void* param, float reps, float aeps, float* line); points count disType param

Array of 3D points. Number of points. Type of the distance used to fit the data to a line. Pointer to a user-defined function that calculates the weights for the type CV_DIST_USER or the pointer to a float user-defined metric parameter c for the Fair and Welsch distance types. Used for iteration stop criteria. If zero, the default value of 0.01 is used. Pointer to the array of 6 floats. When the function exits, the first three elements contain the direction vector of the line normalized to 1, the other three contain coordinates of a point that belongs to the line.

reps, aeps line

Discussion The function FitLine3D fits a 3D line to a set of points on the plane. Possible distance type values are listed below.
CV_DIST_L2 CV_DIST_L1 CV_DIST_L12 CV_DIST_FAIR CV_DIST_WELSCH CV_DIST_USER c =1.3998.

Standard least squares ( x )

= x

c x ( x ) = ----- 1 exp -c 2

,c = 2.9846.

Uses a user-defined function to calculate the weight. The parameter param should point to the function.

11-15

OpenCV Reference Manual

Structural Analysis Reference

The line equation is [ V ( r r 0 ) ] = ( l ine [ 3 ] , line [ 4 ] , line [ 5 . In this algorithm

W ( d ( r i ) ) ri r0

= 0,

where =

( l ine [ 0 ] , line [ 1 ] , line [ 2 V = 1 ,

and

is the mean of the input vectors with weights, that is,

r 0 = -------------------------------------. W ( d ( ri ) )
i

The parameters reps and aeps are iteration thresholds. If the distance between two values of r 0 calculated from two iterations is less than the value of the parameter reps, (the distance type CV_DIST_C is used in this case) and the angle in radians between two vectors V is less than the parameter aeps, then the iteration is stopped. The specification for the user-defined weight function is
void userWeight ( float* dist, int count, float* w ); dist count w

The function should fill the weights array with values of weights calculated from 1 d distance valuesw [ i ] = f ( d [ i ] ) The function f ( x ) = --- ------- has to be monotone decreasing. .
x dx

Project3D
Projects array of 3D points to coordinate axis.
void cvProject3D ( CvPoint3D32f* int xindx, int yindx); points3D count points2D xindx points3D, int count, CvPoint2D32f* points2D,

Pointer to the array of distance values. Number of elements. Pointer to the output array of weights.

Source array of 3D points. Number of points. Target array of 2D points. Index of the 3D coordinate from 0 to 2 that is to be used as x-coordinate.

11-16

OpenCV Reference Manual

Structural Analysis Reference

yindx

Index of the 3D coordinate from 0 to 2 that is to be used as y-coordinate.

Discussion The function Project3D used with the function PerspectiveProject is intended to provide a general way of projecting a set of 3D points to a 2D plane. The function copies two of the three coordinates specified by the parameters xindx and yindx of each 3D point to a 2D points array.

ConvexHull
Finds convex hull of points set.
void cvConvexHull( CvPoint* points, int numPoints, CvRect* boundRect, int orientation, int* hull, int* hullsize ); points numPoints boundRect orientation

Pointer to the set of 2D points. Number of points. Pointer to the bounding rectangle of points set; not used. Output order of the convex hull vertices CV_CLOCKWISE or CV_COUNTER_CLOCKWISE. Indices of convex hull vertices in the input array. Number of vertices in convex hull; output parameter.

hull hullsize

Discussion The function ConvexHull takes an array of points and puts out indices of points that are convex hull vertices. The function uses Quicksort algorithm for points sorting. The standard, that is, bottom-left XY coordinate system, is used to define the order in which the vertices appear in the output array.

11-17

OpenCV Reference Manual

Structural Analysis Reference

ContourConvexHull
Finds convex hull of points set.
CvSeq* cvContourConvexHull( CvSeq* contour, int orientation, CvMemStorage* storage ); contour orientation

Sequence of 2D points. Output order of the convex hull vertices CV_CLOCKWISE or CV_COUNTER_CLOCKWISE. Memory storage where the convex hull must be allocated.

storage

Discussion The function ContourConvexHull takes an array of points and puts out indices of points that are convex hull vertices. The function uses Quicksort algorithm for points sorting. The standard, that is, bottom-left XY coordinate system, defines the order in which the vertices appear in the output array. The function returns CvSeq that is filled with pointers to those points of the source contour that belong to the convex hull.

ConvexHullApprox
Finds approximate convex hull of points set.
void cvConvexHullApprox( CvPoint* points, int numPoints, CvRect* boundRect, int bandWidth,int orientation, int* hull, int* hullsize ); points numPoints boundRect bandWidth

Pointer to the set of 2D points. Number of points. Pointer to the bounding rectangle of points set; not used. Width of band used by the algorithm.

11-18

11-21

OpenCV Reference Manual

Structural Analysis Reference

MinAreaRect
Finds circumscribed rectangle of minimal area for given convex contour.
void cvMinAreaRect ( CvPoint* points, int n, int left, int bottom, int right, int top, CvPoint2D32f* anchor, CvPoint2D32f* vect1, CvPoint2D32f* vect2 ); points n left bottom right top anchor vect1 vect2

Sequence of convex polygon points. Number of input points. Index of the extreme left point. Index of the extreme bottom point. Index of the extreme right point. Index of the extreme top point. Pointer to one of the output rectangle corners. Pointer to the vector that represents one side of the output rectangle. Pointer to the vector that represents another side of the output rectangle.

11-22

OpenCV Reference Manual

Structural Analysis Reference

Discussion The function MinAreaRect returns a circumscribed rectangle of the minimal area. The output parameters of this function are the corner of the rectangle and two incident edges of the rectangle (see Figure 11-3).
Figure 11-3 Minimal Area Bounding Rectangle

CalcPGH
Calculates pair-wise geometrical histogram for contour.
void cvCalcPGH( CvSeq* contour, CvHistogram* hist ); contour hist

Input contour. Calculated histogram; must be two-dimensional.

Discussion The function CalcPGH calculates a pair-wise geometrical histogram for the contour. The algorithm considers every pair of the contour edges. The angle between the edges and the minimum/maximum distances are determined for every pair. To do this each of the edges in turn is taken as the base, while the function loops through all the other edges. When the base edge and any other edge are considered, the minimum and

11-23

OpenCV Reference Manual

Structural Analysis Reference

maximum distances from the points on the non-base edge and line of the base edge are selected. The angle between the edges defines the row of the histogram in which all the bins that correspond to the distance between the calculated minimum and maximum distances are incremented. The histogram can be used for contour matching.

MinEnclosingCircle
Finds minimal enclosing circle for 2D-point set.
void cvFindMinEnclosingCircle( CvSeq* seq, CvPoint2D32f* center, float* radius ); seq

Sequence that contains the input point set. Only points with integer coordinates (CvPoint) are supported. Output parameter. The center of the enclosing circle. Output parameter. The radius of the enclosing circle.

center radius

Discussion The function FindMinEnclosingCircle finds the minimal enclosing circle for the planar point set. Enclosing means that all the points from the set are either inside or on the boundary of the circle. Minimal means that there is no enclosing circle of a smaller radius.

Contour Processing Data Types

The OpenCV Library functions use special data structures to represent the contours and contour binary tree in memory, namely the structures CvSeq and CvContourTree. Below follows the definition of the structure CvContourTree in the C language.
Example 11-1 CvContourTree
typedef struct CvContourTree { CV_SEQUENCE_FIELDS() CvPoint p1; /*the start point of the binary tree root*/ CvPoint p2; /*the end point of the binary tree

11-24

OpenCV Reference Manual

Structural Analysis Reference

Example 11-1 CvContourTree (continued)

root*/ } CvContourTree;

Geometry Data Types

Example 11-2 CvConvexityDefect
typedef struct { CvPoint* start; CvPoint* end; CvPoint* depth_point; float depth; } CvConvexityDefect;

//start point of defect //end point of defect //fathermost point //depth of defect

11-25

OpenCV Reference Manual

Structural Analysis Reference

11-26

Object Recognition Reference

Table 12-1 Image Recognition Functions and Data Types
Group Eigen Objects Functions Function Name Functions

12
Description Calculates a covariance matrix of the input objects group using previously calculated averaged object. Calculates orthonormal eigen basis and the averaged object for a group of the input objects. Calculates one decomposition coefficient of the input object using the previously calculated eigen object and the averaged object. Calculates all decomposition coefficients for the input object. Calculates an object projection to the eigen sub-space.

CalcCovarMatrixEx

CalcEigenObjects

CalcDecompCoeff

EigenDecomposite

EigenProjection

12-1

OpenCV Reference Manual

Object Recognition Reference

Table 12-1 Image Recognition Functions and Data Types (continued)

Group Embedded Hidden Markov Models Functions Function Name Description Creates a 2D embedded HMM. Frees all the memory used by HMM. Creates new structures to store image observation vectors. Frees all memory used by observations and clears pointer to the structure CvImgObsInfo. Extracts observation vectors from the image. Performs uniform segmentation of image observations by HMM states. Segments all observations within every internal state of HMM by state mixture components. Estimates all parameters of every HMM state. Computes transition probability matrices for embedded HMM. Computes probability of every observation of several images. Executes Viterbi algorithm for embedded HMM.

Create2DHMM Release2DHMM CreateObsInfo

ReleaseObsInfo

ImgToObs_DCT UniformImgSegm

InitMixSegm

EstimateHMMStateParams

EstimateTransProb

EstimateObsProb

EViterbi

12-2

OpenCV Reference Manual

Object Recognition Reference

Table 12-1 Image Recognition Functions and Data Types (continued)

Group Function Name Description Segments observations from all training images by mixture components of newly Viterbi algorithm-assigned states. Shows the use of the function when the size of free RAM is sufficient for all input and eigen objects allocation. Shows the use of the function when all objects and/or eigen objects cannot be allocated in free RAM. Represents 1D HMM and 2D embedded HMM models. Represents image observations.

MixSegmL2

Data Types Use of Eigen Object Functions Use of Function cvCalcEigenObjects in Direct Access Mode

User Data Structure, I/O Callback Functions, and Use of Function cvCalcEigenObjects in Callback Mode HMM Structures Embedded HMM Structure

Image Observation Structure

Eigen Objects Functions

CalcCovarMatrixEx
Calculates covariance matrix for group of input objects.
void cvCalcCovarMatrixEx( int nObjects, void* input, int ioFlags, int ioBufSize, uchar* buffer, void* userData, IplImage* avg, float* covarMatrix );

12-3

OpenCV Reference Manual

Object Recognition Reference

nObjects input

Number of source objects. Pointer either to the array of IplImage input objects or to the read callback function according to the value of the parameter ioFlags. Input/output flags. Input/output buffer size. Pointer to the input/output buffer. Pointer to the structure that contains all necessary data for the callback functions. Averaged object. Covariance matrix. An output parameter; must be allocated before the call.

ioFlags ioBufSize buffer userData

avg covarMatrix

Discussion The function CalcCovarMatrixEx calculates a covariance matrix of the input objects group using previously calculated averaged object. Depending on ioFlags parameter it may be used either in direct access or callback mode. If ioFlags is not CV_EIGOBJ_NO_CALLBACK, buffer must be allocated before calling the function .

CalcEigenObjects
Calculates orthonormal eigen basis and averaged object for group of input objects.
void cvCalcEigenObjects ( int nObjects, void* input, void* output, int ioFlags, int ioBufSize, void* userData, CvTermCriteria* calcLimit, IplImage* avg, float* eigVals; nObjects input

Number of source objects. Pointer either to the array of IplImage input objects or to the read callback function according to the value of the parameter ioFlags. Pointer either to the array of eigen objects or to the write callback function according to the value of the parameter ioFlags.

output

12-4

OpenCV Reference Manual

Object Recognition Reference

ioFlags ioBufSize userData

Input/output flags. Input/output buffer size in bytes. The size is zero, if unknown. Pointer to the structure that contains all necessary data for the callback functions. Criteria that determine when to stop calculation of eigen objects. Averaged object. Pointer to the eigenvalues array in the descending order; may be NULL.

calcLimit avg eigVals

Discussion The function CalcEigenObjects calculates orthonormal eigen basis and the averaged object for a group of the input objects. Depending on ioFlags parameter it may be used either in direct access or callback mode. Depending on the parameter calcLimit, calculations are finished either after first calcLimit.maxIters dominating eigen objects are retrieved or if the ratio of the current eigenvalue to the largest eigenvalue comes down to calcLimit.epsilon threshold. The value calcLimit->type must be CV_TERMCRIT_NUMB, CV_TERMCRIT_EPS, or CV_TERMCRIT_NUMB | CV_TERMCRIT_EPS. The function returns the real values calcLimit->maxIter and calcLimit->epsilon. The function also calculates the averaged object, which must be created previously. Calculated eigen objects are arranged according to the corresponding eigenvalues in the descending order. The parameter eigVals may be equal to NULL, if eigenvalues are not needed. The function CalcEigenObjects uses the functionCalcCovarMatrixEx.

CalcDecompCoeff
Calculates decomposition coefficient of input object.
double cvCalcDecompCoeff( IplImage* obj, IplImage* eigObj, IplImage* avg );

12-5

OpenCV Reference Manual

Object Recognition Reference

obj eigObj avg

Input object. Eigen object. Averaged object.

Discussion The function CalcDecompCoeff calculates one decomposition coefficient of the input object using the previously calculated eigen object and the averaged object.

EigenDecomposite
Calculates all decomposition coefficients for input object.
void cvEigenDecomposite( IplImage* obj, int nEigObjs, void* eigInput, int ioFlags, void* userData, IplImage* avg, float* coeffs ); obj nEigObjs eigInput

Input object. Number of eigen objects. Pointer either to the array of IplImage input objects or to the read callback function according to the value of the parameter ioFlags. Input/output flags. Pointer to the structure that contains all necessary data for the callback functions. Averaged object. Calculated coefficients; an output parameter.

ioFlags userData

avg coeffs

Discussion The function EigenDecomposite calculates all decomposition coefficients for the input object using the previously calculated eigen objects basis and the averaged object. Depending on ioFlags parameter it may be used either in direct access or callback mode.

12-6

OpenCV Reference Manual

Object Recognition Reference

EigenProjection
Calculates object projection to the eigen sub-space.
void cvEigenProjection ( int nEigObjs, void* eigInput, int ioFlags, void* userData, float* coeffs, IplImage* avg, IplImage* proj ); nEigObjs eigInput

Number of eigen objects. Pointer either to the array of IplImage input objects or to the read callback function according to the value of the parameter ioFlags. Input/output flags. Pointer to the structure that contains all necessary data for the callback functions. Previously calculated decomposition coefficients. Averaged object. Decomposed object projection to the eigen sub-space.

ioFlags userData

coeffs avg proj

OpenCV Reference Manual

obsSize

delta

12-14

OpenCV Reference Manual

Object Recognition Reference

Discussion The function ImgToObs_DCT extracts observation vectors, that is, DCT coefficients, from the image. The user must pass obsInfo.obs as the parameter obs to use this function with other HMM functions and use the structure obsInfo of the CvImgObsInfo type.
Example 12-3 Calculating Observations for HMM
CvImgObsInfo* obs_info; ... cvImgToObs_DCT( image,obs_info->obs, //!!! dctSize, obsSize, delta );

UniformImgSegm
Performs uniform segmentation of image observations by HMM states.
void cvUniformImgSegm( CvImgObsInfo* obsInfo, CvEHMM* hmm); obsInfo hmm

Observations structure. HMM structure.

12-15

OpenCV Reference Manual

Array of pointers to the observation structures. Length of the array. HMM.

Discussion The function MixSegmL2 segments observations from all training images by mixture components of newly Viterbi algorithm-assigned states. The function uses Euclidean distance to group vectors around the existing mixtures centers.

HMM Structures
In order to support embedded models the user must define structures to represent 1D HMM and 2D embedded HMM model.
Example 12-4 Embedded HMM Structure
typedef struct _CvEHMM { int level; int num_states; float* transP; float** obsProb; union { CvEHMMState* state; struct _CvEHMM* ehmm; } u; }CvEHMM;

Below is the description of the CvEHMM fields:

12-19

OpenCV Reference Manual

Object Recognition Reference

level

Level of embedded HMM. If level==0, HMM is most external. In 2D HMM there are two types of HMM: 1 external and several embedded. External HMM has level==1, embedded HMMs have level==0. Number of states in 1D HMM. State-to-state transition probability, square matrix ( num _ state num _ state ). Observation probability matrix. Array of HMM states. For the last-level HMM, that is, an HMM without embedded HMMs, HMM states are real. Array of embedded HMMs. If HMM is not last-level, then HMM states are not real and they are HMMs.

num_states transP

obsProb state

ehmm

For representation of observations the following structure is defined:

Example 12-5 Image Observation Structure
typedef struct CvImgObsInfo { int obs_x; int obs_y; int obs_size; float** obs; int* state; int* mix; }CvImgObsInfo;

This structure is used for storing observation vectors extracted from 2D image.
obs_x obs_y obs_size obs

Number of observations in the horizontal direction. Number of observations in the vertical direction. Length of every observation vector. Pointer to observation vectors stored consequently. Number of vectors is obs_x*obs_y. Array of indices of states, assigned to every observation vector. Index of mixture component, corresponding to the observation vector within an assigned state.

state mix

12-20

OpenCV Reference Manual

Object Recognition Reference

12-21

OpenCV Reference Manual

Object Recognition Reference

12-22

3D Reconstruction Reference
Table 13-1 3D Reconstruction Functions
Group Camera Calibration Functions Function Name

13
Description Calibrates the camera with single precision. Calibrates camera with double precision. Finds the extrinsic camera parameters for the pattern. Finds extrinsic camera parameters for the pattern with double precision. Converts the rotation matrix to the rotation vector and vice versa with single precision. Converts the rotation matrix to the rotation vector or vice versa with double precision. Corrects camera lens distortion in the case of a single image. Calculates arrays of distorted points indices and interpolation coefficients.

CalibrateCamera CalibrateCamera_64d FindExtrinsicCameraParams

FindExtrinsicCameraParams_64d

Rodrigues

Rodrigues_64d

UnDistortOnce

UnDistortInit

13-1

OpenCV Reference Manual

3D Reconstruction Reference

Table 13-1 3D Reconstruction Functions (continued)

Group Function Name Description Corrects camera lens distortion using previously calculated arrays of distorted points indices and undistortion coefficients. Finds approximate positions of internal corners of the chessboard. Calculates the fundamental matrix from several pairs of correspondent points in images from two cameras. Calculates scanlines coordinates for two cameras by fundamental matrix. Rectifies the image so that the scanlines in the rectified image are horizontal. Retrieves scanlines from the rectified image and breaks each scanline down into several runs. Finds correspondence between two sets of runs of two warped images. Finds coordinates of scanlines for the virtual camera with the given camera position.

UnDistort

FindChessBoardCornerGuesses

View Morphing Functions

FindFundamentalMatrix

MakeScanlines

PreWarpImage

FindRuns

DynamicCorrespondMulti

MakeAlphaScanlines

13-2

OpenCV Reference Manual

3D Reconstruction Reference

Table 13-1 3D Reconstruction Functions (continued)

Group Function Name Description Morphs two pre-warped images using information about stereo correspondence. Warps the rectified morphed image back. Deletes moire from the given image. Allocates memory for the object structure and computes the object inverse matrix. Implements POSIT algorithm. Deallocates the 3D object structure. Finds an arm region in the 3D range image data. Finds an arm region in the 3D range image data and defines the arm orientation. Creates an arm mask on the image plane. Calculates the homograph matrix for the initial image transformation. Calculates the arm mask probability density from the two 2D histograms. Calculates the maximum rectangle for two input rectangles.

MorphEpilinesMulti

PostWarpImage DeleteMoire
POSIT Functions

CreatePOSITObject

POSIT ReleasePOSITObject FindHandRegion

FindHandRegionA

CreateHandMask CalcImageHomography

CalcProbDensity

MaxRect

13-3

OpenCV Reference Manual

3D Reconstruction Reference

Camera Calibration Functions

CalibrateCamera
Calibrates camera with single precision.
void cvCalibrateCamera( int numImages, int* numPoints, CvSize imageSize, CvPoint2D32f* imagePoints32f, CvPoint3D32f* objectPoints32f, CvVect32f distortion32f, CvMatr32f cameraMatrix32f, CvVect32f transVects32f, CvMatr32f rotMatrs32f, int useIntrinsicGuess); numImages numPoints imageSize imagePoints32f objectPoints32f distortion32f cameraMatrix32f transVects32f

Number of the images. Array of the number of points in each image. Size of the image. Pointer to the images. Pointer to the pattern. Array of four distortion coefficients found. Camera matrix found. Array of translate vectors for each pattern position in the image. Array of the rotation matrix for each pattern position in the image. Intrinsic guess. If equal to 1, intrinsic guess is needed.

rotMatrs32f

useIntrinsicGuess

Discussion The function CalibrateCamera calculates the camera parameters using information points on the pattern object and pattern object images.

13-4

OpenCV Reference Manual

3D Reconstruction Reference

CalibrateCamera_64d
Calibrates camera with double precision.
void cvCalibrateCamera_64d( int numImages, int* numPoints, CvSize imageSize, CvPoint2D64d* imagePoints, CvPoint3D64d* objectPoints, CvVect64d distortion, CvMatr64d cameraMatrix, CvVect64d transVects, CvMatr64d rotMatrs, int useIntrinsicGuess); numImages numPoints imageSize imagePoints objectPoints distortion cameraMatrix transVects

Number of the images. Array of the number of points in each image. Size of the image. Pointer to the images. Pointer to the pattern. Distortion coefficients found. Camera matrix found. Array of the translate vectors for each pattern position on the image. Array of the rotation matrix for each pattern position on the image. Intrinsic guess. If equal to 1, intrinsic guess is needed.

rotMatrs

useIntrinsicGuess

Discussion The function CalibrateCamera_64d is basically the same as the function CalibrateCamera, but uses double precision.

13-5

OpenCV Reference Manual

3D Reconstruction Reference

FindExtrinsicCameraParams
Finds extrinsic camera parameters for pattern.
void cvFindExtrinsicCameraParams( int numPoints, CvSize imageSize, CvPoint2D32f* imagePoints32f, CvPoint3D32f* objectPoints32f, CvVect32f focalLength32f, CvPoint2D32f principalPoint32f, CvVect32f distortion32f, CvVect32f rotVect32f, CvVect32f transVect32f); numPoints ImageSize imagePoints32f objectPoints32f focalLength32f principalPoint32f distortion32f rotVect32f transVect32f

Number of the points. Size of the image. Pointer to the image. Pointer to the pattern. Focal length. Principal point. Distortion. Rotation vector. Translate vector.

Discussion The function FindExtrinsicCameraParams finds the extrinsic parameters for the pattern.

13-6

OpenCV Reference Manual

3D Reconstruction Reference

Distortion data array. Interpolation toggle (optional).

Discussion The function UnDistortInit calculates arrays of distorted points indices and interpolation coefficients using known matrix of the camera intrinsic parameters and distortion coefficients. It must be used before calling the function UnDistort. Matrix of the camera intrinsic parameters and distortion coefficients k1, p2 must be preliminarily calculated by the function CalibrateCamera.
k2, p1, and

The data array must be allocated in the main function before use of the function UnDistortInit. If interpolate = 0, its length must be size.width*size.height elements; otherwise 3*size.width*size.height elements. If interpolate = 0, inter-pixel interpolation is disabled; otherwise default bilinear interpolation is used.

UnDistort
Corrects camera lens distortion.
void cvUnDistort ( IplImage* srcImage, IplImage* dstImage, int* data, int interpolate=1 ); srcImage dstImage data interpolate

Source (distorted) image. Destination (corrected) image. Distortion data array. Interpolation toggle (optional).

Discussion The function UnDistort corrects camera lens distortion using previously calculated arrays of distorted points indices and undistortion coefficients. It is used if a sequence of frames must be corrected.

13-10

OpenCV Reference Manual

3D Reconstruction Reference

Preliminarily, the function UnDistortInit calculates the array data. If interpolate = 0, then inter-pixel interpolation is disabled; otherwise bilinear interpolation is used. In the latter case the function acts slower, but quality of the corrected image increases.

FindChessBoardCornerGuesses
Finds approximate positions of internal corners of the chessboard.
int cvFindChessBoardCornerGuesses(IplImage* img, IplImage* thresh, CvSize etalonSize, CvPoint2D32f* corners, int* cornerCount ); img thresh etalonSize

Source chessboard view; must have the depth of IPL_DEPTH_8U. Temporary image of the same size and format as the source image. Number of inner corners per chessboard row and column. The width (the number of columns) must be less or equal to the height (the number of rows). For chessboard see Figure 6-1. Pointer to the corner array found. Signed value whose absolute value is the number of corners found. A positive number means that a whole chessboard has been found and a negative number means that not all the corners have been found.

corners cornerCount

Discussion The function FindChessBoardCornerGuesses attempts to determine whether the input image is a view of the chessboard pattern and locate internal chessboard corners. The function returns non-zero value if all the corners have been found and they have been placed in a certain order (row by row, left to right in every row), otherwise, if the function fails to find all the corners or reorder them, the function returns 0. For example, a simple chessboard has 8x8 squares and 7x7 internal corners, that is, points, where the squares are tangent. The word approximate in the above description

13-11

OpenCV Reference Manual

3D Reconstruction Reference

means that the corner coordinates found may differ from the actual coordinates by a couple of pixels. To get more precise coordinates, the user may use the function FindCornerSubPix.

View Morphing Functions

FindFundamentalMatrix
Calculates fundamental matrix from several pairs of correspondent points in images from two cameras.
void cvFindFundamentalMatrix( int* points1, int* points2, int numpoints, int method, CvMatrix3* matrix); points1 points2 numpoints method

Pointer to the array of correspondence points in the first image. Pointer to the array of correspondence points in the second image. Number of the point pairs. Method for finding the fundamental matrix; currently not used, must be zero. Resulting fundamental matrix.

matrix

Discussion The function FindFundamentalMatrix finds the fundamental matrix for two cameras from several pairs of correspondent points in images from the cameras. If the number of pairs is less than 8 or the points lie very close to each other or on the same planar surface, the matrix is calculated incorrectly.

13-12

OpenCV Reference Manual

3D Reconstruction Reference

MakeScanlines
Calculates scanlines coordinates for two cameras by fundamental matrix.
void cvMakeScanlines( CvMatrix3* matrix, CvSize imgSize, int* scanlines1, int* scanlines2, int* lens1, int* lens2, int* numlines); matrix imgSize scanlines1 scanlines2 lens1

Fundamental matrix. Size of the image. Pointer to the array of calculated scanlines of the first image. Pointer to the array of calculated scanlines of the second image. Pointer to the array of calculated lengths (in pixels) of the first image scanlines. Pointer to the array of calculated lengths (in pixels) of the second image scanlines. Pointer to the variable that stores the number of scanlines.

lens2

numlines

Discussion The function MakeScanlines finds coordinates of scanlines for two images. This function returns the number of scanlines. The function does nothing except calculating the number of scanlines if the pointers scanlines1 or scanlines2 are equal to zero.

PreWarpImage
Rectifies image.
void cvPreWarpImage( int numLines, IplImage* img, uchar* dst, int* int* scanlines); numLines dstNums,

Number of scanlines for the image.

Discussion The function MorphEpilinesMulti morphs two pre-warped images using information about correspondence between the scanlines of two images.

PostWarpImage
Warps rectified morphed image back.
void cvPostWarpImage( int numLines, uchar* src, int* srcNums, IplImage* img, int* scanlines); numLines

Number of the scanlines.

13-17

OpenCV Reference Manual

3D Reconstruction Reference

src srcNums img scanlines

Pointer to the prewarp image virtual image. Number of the scanlines in the image. Resulting unwarp image. Pointer to the array of scanlines data.

Discussion The function PostWarpImage warps the resultant image from the virtual camera by storing its rows across the scanlines whose coordinates are calculated by MakeAlphaScanlines function.

DeleteMoire
Deletes moire in given image.
void cvDeleteMoire( IplImage* img); img

Image.

Discussion The function DeleteMoire deletes moire from the given image. The post-warped image may have black (un-covered) points because of possible holes between neighboring scanlines. The function deletes moire (black pixels) from the image by substituting neighboring pixels for black pixels. If all the scanlines are horizontal, the function may be omitted.

13-18

OpenCV Reference Manual

3D Reconstruction Reference

POSIT Functions

CreatePOSITObject
Initializes structure containing object information.
CvPOSITObject* cvCreatePOSITObject( CvPoint3D32f* points, int numPoints ); points numPoints

Pointer to the points of the 3D object model. Number of object points.

Discussion The function CreatePOSITObject allocates memory for the object structure and computes the object inverse matrix. The preprocessed object data is stored in the structure CvPOSITObject, internal for OpenCV, which means that the user cannot directly access the structure data. The user may only create this structure and pass its pointer to the function. Object is defined as a set of points given in a coordinate system. The function POSIT computes a vector that begins at a camera-related coordinate system center and ends at the points[0] of the object. Once the work with a given object is finished, the function ReleasePOSITObject must be called to free memory.

POSIT
Implements POSIT algorithm.
void cvPOSIT( CvPoint2D32f* imagePoints, CvPOSITObject* pObject, double focalLength, CvTermCriteria criteria, CvMatrix3* rotation, CvPoint3D32f* translation);

13-19

OpenCV Reference Manual

3D Reconstruction Reference

imagePoints pObject focalLength criteria rotation translation

Pointer to the object points projections on the 2D image plane. Pointer to the object structure. Focal length of the camera used. Termination criteria of the iterative POSIT algorithm. Matrix of rotations. Translation vector.

Discussion The function POSIT implements POSIT algorithm. Image coordinates are given in a camera-related coordinate system. The focal length may be retrieved using camera calibration functions. At every iteration of the algorithm new perspective projection of estimated pose is computed. Difference norm between two projections is the maximal distance between corresponding points. The parameter criteria.epsilon serves to stop the algorithm if the difference is small.

ReleasePOSITObject
Deallocates 3D object structure.
void cvReleasePOSITObject( CvPOSITObject** ppObject ); ppObject

Address of the pointer to the object structure.

Discussion The function ReleasePOSITObject releases memory previously allocated by the function CreatePOSITObject.

13-20

OpenCV Reference Manual

3D Reconstruction Reference

Gesture Recognition Functions

FindHandRegion
Finds arm region in 3D range image data.
void cvFindHandRegion( CvPoint3D32f* points, int count, CvSeq* indexs, float* line, CvSize2D32f size, int flag, CvPoint3D32f* center, CvMemStorage* storage, CvSeq** numbers); points count indexs line size flag center storage numbers

Pointer to the input 3D point data. Numbers of the input points. Sequence of the input points indices in the initial image. Pointer to the input points approximation line. Size of the initial image. Flag of the arm orientation. Pointer to the output arm center. Pointer to the memory storage. Pointer to the output sequence of the points indices.

Discussion The function FindHandRegion finds the arm region in 3D range image data. The coordinates of the points must be defined in the world coordinates system. Each input point has user-defined transform indices indexs in the initial image. The function finds the arm region along the approximation line from the left, if flag = 0, or from the right, if flag = 1, in the points maximum accumulation by the points projection histogram calculation. Also the function calculates the center of the arm region and the indices of the points that lie near the arm center. The function FindHandRegion assumes that the arm length is equal to about 0.25m in the world coordinate system.

13-21

OpenCV Reference Manual

3D Reconstruction Reference

FindHandRegionA
Finds arm region in 3D range image data and defines arm orientation.
void cvFindHandRegionA( CvPoint3D32f* points, int count, CvSeq* indexs, float* line, CvSize2D32f size, int jCenter, CvPoint3D32f* center, CvMemStorage* storage, CvSeq** numbers); points count indexs line size jCenter center storage numbers

Pointer to the input 3D point data. Number of the input points. Sequence of the input points indices in the initial image. Pointer to the input points approximation line. Size of the initial image. Input j-index of the initial image center. Pointer to the output arm center. Pointer to the memory storage. Pointer to the output sequence of the points indices.

Discussion The function FindHandRegionA finds the arm region in the 3D range image data and defines the arm orientation (left or right). The coordinates of the points must be defined in the world coordinates system. The input parameter jCenter is the index j of the initial image center in pixels (width/2). Each input point has user-defined transform indices on the initial image (indexs). The function finds the arm region along approximation line from the left or from the right in the points maximum accumulation by the points projection histogram calculation. Also the function calculates the center of the arm region and the indices of points that lie near the arm center. The function FindHandRegionA assumes that the arm length is equal to about 0.25m in the world coordinate system.

13-22

OpenCV Reference Manual

3D Reconstruction Reference

CreateHandMask
Creates arm mask on image plane.
void cvCreateHandMask(CvSeq* numbers, IplImage *imgMask, CvRect *roi); numbers imgMask roi

Sequence of the input points indices in the initial image. Pointer to the output image mask. Pointer to the output arm ROI.

Discussion The function CreateHandMask creates an arm mask on the image plane. The pixels of the resulting mask associated with the set of the initial image indices indexs associated with hand region have the maximum unsigned char value (255). All remaining pixels have the minimum unsigned char value (0). The output image mask imgMask has to have the IPL_DEPTH_8U type and the number of channels is 1.

CalcImageHomography
Calculates homography matrix.
void cvCalcImageHomography(float* line, CvPoint3D32f* center, float intrinsic[3][3], float homography[3][3]); line center intrinsic homography

Pointer to the input 3D line. Pointer to the input arm center. Matrix of the intrinsic camera parameters. Output homography matrix.

13-23

OpenCV Reference Manual

3D Reconstruction Reference

Discussion The function CalcImageHomography calculates the homograph matrix for the initial image transformation from image plane to the plane, defined by 3D arm line (See Figure 6-10 in Programmer Guide 3D Reconstruction Chapter). If n1=(nx,ny)and n2=(nx,nz) are coordinates of the normals of the 3D line projection of planes XY and XZ, then the resulting image homography matrix is calculated as 1 H = A ( R h + ( I 3 3 R h ) x h [ 0, 0, 1 ] ) A , where Rh is the 3x3 matrix R h = R 1 R 2 , and ,
Th T T R 1 = [ n 1 u z, n 1, u z ], R 2 = [ u y n 2, u y, n 2 ], uz = [ 0, 0, 1 ] , u y = [ 0, 1, 0 ] , x h = ----- = Tz Tx T y ----- , ----- , 1 - Tz T z
T

where ( T x, T y , T z ) is the arm center coordinates in the world coordinate system, and A is the intrinsic camera parameters matrix
fx 0 cx A =

0 fy cy 0 0 1

The diagonal entries fx and fy are the camera focal length in units of horizontal and vertical pixels and the two remaining entries c x, c y are the principal point image coordinates.

CalcProbDensity
Calculates arm mask probability density on image plane.
void cvCalcProbDensity (CvHistogram* hist, CvHistogram* histMask, CvHistogram* histDens ); hist histMask histDens

Input image histogram. Input image mask histogram. Resulting probability density histogram.

13-24

OpenCV Reference Manual

3D Reconstruction Reference

Discussion The function CalcProbDensity calculates the arm mask probability density from the two 2D histograms. The input histograms have to be calculated in two channels on the initial image. If { h ij } and { hm ij }, 1 i B i, 1 j B j are input histogram and mask histogram respectively, then the resulting probability density histogram p ij is calculated as
m ij --------- 255, if hij 0, h ij

p ij =

0, if hij = 0, 255, if m ij > h ij

So the values of the

p ij

are between 0 and 255.

MaxRect
Calculates the maximum rectangle.
void cvMaxRect (CvRect* rect1, CvRect* rect2, CvRect* maxRect ); rect1 rect2 maxRect

First input rectangle. Second input rectangle. Resulting maximum rectangle.

13-25

Table 14-1 Basic Structures and Operations Functions, Macros, and Data Types (continued)
Name Description Creates an empty set with a specified header size. Adds an element to the set. Removes an element from the set. Finds a set element by index. Empties the set. Creates an empty graph. Adds a vertex to the graph. Removes a vertex from the graph. Removes a vertex from the graph together with all the edges incident to it. Adds an edge to the graph. Adds an edge to the graph given the starting and the ending vertices. Removes an edge from the graph. Removes an edge from the graph that connects given vertices. Finds the graph edge that connects given vertices. Finds the graph edge that connects given vertices. Finds an edge in the graph. Counts the edges incident to the graph vertex, both incoming and outcoming, and returns the result. Removes all the vertices and edges from the graph. Finds the graph vertex by index. Returns the index of the graph vertex. Returns the index of the graph edge.

CreateSet SetAdd SetRemove GetSetElem ClearSet CreateGraph GraphAddVtx GraphRemoveVtx GraphRemoveVtxByPtr GraphAddEdge GraphAddEdgeByPtr GraphRemoveEdge GraphRemoveEdgeByPtr FindGraphEdge FindGraphEdgeByPtr GraphVtxDegree GraphVtxDegreeByPtr ClearGraph GetGraphVtx GraphVtxIdx GraphEdgeIdx
Matrix Operations Functions

Alloc AllocArray Free

Allocates memory for the matrix data. Allocates memory for the matrix array data. Frees memory allocated for the matrix data.

14-3

OpenCV Reference Manual

Basic Structures and Operations Reference

Table 14-1 Basic Structures and Operations Functions, Macros, and Data Types (continued)
Name Description Frees memory allocated for the matrix array data. Computes sum of two matrices. Computes difference of two matrices. Multiplies every element of the matrix by a scalar. Calculates dot product of two vectors in Euclidian metrics. Calculates the cross product of two 3D vectors. Multiplies matrices. Calculates the product of a matrix and its transposition. Transposes a matrix. Inverts a matrix. Returns the trace of a matrix. Returns the determinant of a matrix. Copies one matrix to another. Sets the matrix to zero. Sets the matrix to identity. Calculates the weighted distance between two vectors. Decomposes the source matrix to product of two orthogonal and one diagonal matrices. Computes eigenvalues and eigenvectors of a symmetric matrix. Implements general transform of a 3D vector array.

FreeArray Add Sub Scale DotProduct CrossProduct Mul MulTransposed Transpose Invert Trace Det Copy SetZero SetIdentity Mahalonobis SVD EigenVV PerspectiveProject
Drawing Primitives Functions

Line LineAA Rectangle Circle Ellipse EllipseAA FillPoly

Draws a simple or thick line segment. Draws an antialiased line segment. Draws a simple, thick or filled rectangle. Draws a simple, thick or filled circle. Draws a simple or thick elliptic arc or fills an ellipse sector. Draws an antialiased elliptic arc. Fills an area bounded by several polygonal contours.

14-4

OpenCV Reference Manual

Basic Structures and Operations Reference

Table 14-1 Basic Structures and Operations Functions, Macros, and Data Types (continued)
Name Description Fills convex polygon interior. Draws a set of simple or thick polylines. Draws a set of antialiased polylines. Initializes the font structure. Draws a text string. Retrieves width and height of the text string.

FillConvexPoly PolyLine PolyLineAA InitFont PutText GetTextSize

Utility Functions

AbsDiff AbsDiffS MatchTemplate CvtPixToPlane CvtPlaneToPix ConvertScale InitLineIterator SampleLine GetRectSubPix bFastArctan Sqrt bSqrt InvSqrt bInvSqrt bReciprocal bCartToPolar bFastExp

Calculates absolute difference between two images. Calculates absolute difference between an image and a scalar. Fills a specific image for a given image and template. Divides a color image into separate planes. Composes a color image from separate planes. Converts one image to another with linear transformation. Initializes the line iterator and returns the number of pixels between two end points. Reads a raster line to buffer. Retrieves a raster rectangle from the image with sub-pixel accuracy. Calculates fast arctangent approximation for arrays of abscissas and ordinates. Calculates square root of a single argument. Calculates the square root of an array of floats. Calculates the inverse square root of a single float. Calculates the inverse square root of an array of floats. Calculates the inverse of an array of floats. Calculates the magnitude and the angle for an array of abscissas and ordinates. Calculates fast exponent approximation for each element of the input array of floats.

14-5

OpenCV Reference Manual

Basic Structures and Operations Reference

Table 14-1 Basic Structures and Operations Functions, Macros, and Data Types (continued)
Name Description Calculates fast logarithm approximation for each element of the input array. Initializes state of the random number generator. Fills the array with random numbers and updates generator state. Fills the image with a constant value. Changes the range of generated random numbers without reinitializing RNG state. Splits a set of vectors into a given number of clusters. Data Types Memory Storage

bFastLog RandInit bRand FillImage RandSetRange KMeans

CvMemStorage Structure Definition CvMemBlock Structure Definition CvMemStoragePos Structure Definition

Sequence Data

CvSequence Structure Definition

Standard Types of Sequence Elements Standard Kinds of Sequences

Simplifies the extension of the structure CvSeq with additional parameters. Provides definitions of standard sequence elements. Specifies the kind of the sequence. Defines the building block of sequences.

CvSeqBlock Structure Definition

Set Data Structures

CvSet Structure Definition CvSetElem Structure Definition

Graphs Data Structures

CvGraph Structure Definition

Definitions of CvGraphEdge and CvGraphVtx Structures

14-6

OpenCV Reference Manual

Basic Structures and Operations Reference

Table 14-1 Basic Structures and Operations Functions, Macros, and Data Types (continued)
Name Matrix Operations Description

CvMat Structure Definition CvMatArray Structure Definition

Pixel Access

Stores real single-precision or double-precision matrices. Stores arrays of matrices to reduce time call overhead.

CvPixelPosition Structures Definition

Pixel Access Macros CV_INIT_PIXEL_POS CV_MOVE_TO CV_MOVE CV_MOVE_WRAP CV_MOVE_PARAM CV_MOVE_PARAM_WRAP Initializes one of CvPixelPosition structures. Moves to a specified absolute position. Moves by one pixel relative to the current position. Moves by one pixel relative to the current position and wraps when the position reaches the image boundary. Moves by one pixel in a specified direction. Moves by one pixel in a specified direction with wrapping.

Image Functions Reference

CreateImageHeader
Allocates, initializes, and returns structure IplImage.
IplImage* cvCreateImageHeader( CvSize size, int depth, int channels); size depth channels

Image width and height. Image depth. Number of channels.

14-7

OpenCV Reference Manual

Basic Structures and Operations Reference

Discussion The function CreateImageHeader allocates, initializes, and returns the structure IplImage. This call is a shortened form of
iplCreateImageHeader( channels, 0, depth, channels == 1 ? "GRAY" : "RGB", channels == 1 ? "GRAY" : channels == 3 ? "BGR" : "BGRA", IPL_DATA_ORDER_PIXEL, IPL_ORIGIN_TL, 4, size.width, size.height, 0,0,0,0);

CreateImage
Creates header and allocates data.
IplImage* cvCreateImage( CvSize size, int depth, int channels ); size depth channels

Image width and height. Image depth. Number of channels.

Discussion The function CreateImage creates the header and allocates data. This call is a shortened form of
header = cvCreateImageHeader(size,depth,channels); cvCreateImageData(header);

14-8

OpenCV Reference Manual

Basic Structures and Operations Reference

ReleaseImageHeader
Releases header.
void cvReleaseImageHeader( IplImage** image ); image

Double pointer to the deallocated header.

Discussion The function ReleaseImageHeader releases the header. This call is a shortened form of
if( image ) { iplDeallocate( *image, IPL_IMAGE_HEADER | IPL_IMAGE_ROI ); *image = 0; }

ReleaseImage
Releases header and image data.
void cvReleaseImage( IplImage** image ) image

Double pointer to the header of the deallocated image.

Discussion The function ReleaseImage releases the header and the image data. This call is a shortened form of
if( image ) { iplDeallocate( *image, IPL_IMAGE_ALL );

Example 14-1 CvPixelPosition Structures Definition (continued)

pixel line */ topline; /* pointer to the start of the top pixel line */ unsigned char* bottomline; /* pointer to the start of the first line which is below the image */ int x; /* current x coordinate ( in pixels ) */ int width; /* width of the image ( in pixels )*/ int height; /* height of the image ( in pixels )*/ int step; /* distance between lines ( in elements of single plane ) */ int step_arr[3]; /* array: ( 0, -step, step ). It is used for vertical moving */ } CvPixelPosition8u; unsigned char* /*this structure differs from the above only in data type*/ typedef struct _CvPixelPosition8s { char* currline; char* topline; char* bottomline; int x; int width; int height; int step; int step_arr[3]; } CvPixelPosition8s; /* this structure differs from the CvPixelPosition8u only in data type */ typedef struct _CvPixelPosition32f { float* currline; float* topline; float* bottomline; int x; int width; int height; int step; int step_arr[3]; } CvPixelPosition32f;

14-15

OpenCV Reference Manual

Basic Structures and Operations Reference

CV_INIT_PIXEL_POS
Initializes one of CvPixelPosition structures.
#define CV_INIT_PIXEL_POS( pos, origin, step, roi, x, y, orientation ) pos origin step roi x, y orientation

Initialization of structure. Pointer to the left-top corner of ROI. Width of the whole image in bytes. Width and height of ROI. Initial position. Image orientation; could be either
CV_ORIGIN_TL CV_ORIGIN_BL -

top/left orientation, or bottom/left orientation.

CV_MOVE_TO
Moves to specified absolute position.
#define CV_MOVE_TO( pos, x, y, cs ) pos x, y cs

Position structure. Coordinates of the new position. Number of the image channels.

14-16

OpenCV Reference Manual

Basic Structures and Operations Reference

CV_MOVE
Moves by one pixel relative to current position.
#define CV_MOVE_LEFT( pos, cs ) #define CV_MOVE_RIGHT( pos, cs ) #define CV_MOVE_UP( pos, cs ) #define CV_MOVE_DOWN( pos, cs ) #define CV_MOVE_LU( pos, cs ) #define CV_MOVE_RU( pos, cs ) #define CV_MOVE_LD( pos, cs ) #define CV_MOVE_RD( pos, cs ) pos cs

Position structure. Number of the image channels.

CV_MOVE_WRAP
Moves by one pixel relative to current position and wraps when position reaches image boundary.
#define CV_MOVE_LEFT_WRAP( pos, cs ) #define CV_MOVE_RIGHT_WRAP( pos, cs ) #define CV_MOVE_UP_WRAP( pos, cs ) #define CV_MOVE_DOWN_WRAP( pos, cs ) #define CV_MOVE_LU_WRAP( pos, cs ) #define CV_MOVE_RU_WRAP( pos, cs ) #define CV_MOVE_LD_WRAP( pos, cs ) #define CV_MOVE_RD_WRAP( pos, cs ) pos

Position structure.

14-17

OpenCV Reference Manual

Basic Structures and Operations Reference

Number of the image channels.

CV_MOVE_PARAM
Moves by one pixel in specified direction.
#define CV_MOVE_PARAM( pos, shift, cs ) pos cs shift

Position structure. Number of the image channels. Direction; could be any of the following:
CV_SHIFT_NONE, CV_SHIFT_LEFT, CV_SHIFT_RIGHT, CV_SHIFT_UP, CV_SHIFT_DOWN, CV_SHIFT_UL, CV_SHIFT_UR, CV_SHIFT_DL.

CV_MOVE_PARAM_WRAP
Moves by one pixel in specified direction with wrapping.
#define CV_MOVE_PARAM_WRAP( pos, shift, cs ) pos cs shift

Position structure. Number of the image channels. Direction; could be any of the following:

14-18

OpenCV Reference Manual

Basic Structures and Operations Reference

CV_SHIFT_NONE, CV_SHIFT_LEFT, CV_SHIFT_RIGHT, CV_SHIFT_UP, CV_SHIFT_DOWN, CV_SHIFT_UL, CV_SHIFT_UR, CV_SHIFT_DL.

14-19

OpenCV Reference Manual

Basic Structures and Operations Reference

Dynamic Data Structures Reference

Memory Storage Reference
Example 14-2 CvMemStorage Structure Definition
typedef struct CvMemStorage { CvMemBlock* bottom;/* first allocated block */ CvMemBlock* top; /*current memory block - top of the stack */ struct CvMemStorage* parent; /* borrows new blocks from */ int block_size; /* block size */ int free_space; /* free space in the current block */ } CvMemStorage;

Example 14-3 CvMemBlock Structure Definition

Memory storage. New storage top position.

Discussion The function RestoreMemStoragePos restores the position of the storage top from the parameter pos. This function and the function ClearMemStorage are the only methods to release memory occupied in memory blocks. In other words, the occupied space and free space in the storage are continuous. If the user needs to process data and put the result to the storage, there arises a need for the storage space to be allocated for temporary results. In this case the user may simply write all the temporary data to that single storage. However, as a result garbage appears in the middle of the occupied part. See Figure 14-1.

14-23

OpenCV Reference Manual

Basic Structures and Operations Reference

Figure 14-1 Storage Allocation for Temporary Results

Input/Output Storage

Input (Occupied) Data

Input/Output Storage

Temporary Data (Garbage)

Output Data

Saving/Restoring does not work in this case. Creating a child memory storage, however, can resolve this problem. The algorithm writes to both storages simultaneously, and, once done, releases the temporary storage. See Figure 14-2.

14-24

OpenCV Reference Manual

Basic Structures and Operations Reference

Figure 14-2 Release of Temporary Storage

IInput/Output Storage

Temporary Child Storage Will be returned to the parent

Sequence Reference
Example 14-5 CvSequence Structure Definition
#define CV_SEQUENCE_FIELDS() \ int header_size; /* size of sequence header */ \ struct CvSeq* h_prev; /* previous sequence */ \ struct CvSeq* h_next; /* next sequence */ \ struct CvSeq* v_prev; /* 2nd previous sequence */ \ struct CvSeq* v_next; /* 2nd next sequence */ \ int flags; /* micsellaneous flags */ \ int total; /* total number of elements */ \ int elem_size;/* size of sequence element in bytes */ \ char* block_max;/* maximal bound of the last block */ \ char* ptr; /* current write pointer */ \ int delta_elems; /* how many elements allocated when the seq grows */ \ CvMemStorage* storage; /* where the seq is stored */ \ CvSeqBlock* free_blocks; /* free blocks list */ \ CvSeqBlock* first; /* pointer to the first sequence block */ typedef struct CvSeq { CV_SEQUENCE_FIELDS() } CvSeq;

14-25

OpenCV Reference Manual

Basic Structures and Operations Reference

Such an unusual definition simplifies the extension of the structure CvSeq with additional parameters. To extend CvSeq the user may define a new structure and put user-defined fields after all CvSeq fields that are included via the macro CV_SEQUENCE_FIELDS(). The field header_size contains the actual size of the sequence header and must be more than or equal to sizeof(CvSeq). The fields h_prev, h_next, v_prev, v_next can be used to create hierarchical structures from separate sequences. The fields h_prev and h_next point to the previous and the next sequences on the same hierarchical level while the fields v_prev and v_next point to the previous and the next sequence in the vertical direction, that is, parent and its first child. But these are just names and the pointers can be used in a different way. The field first points to the first sequence block, whose structure is described below. The field flags contain miscellaneous information on the type of the sequence and should be discussed in greater detail. By convention, the lowest CV_SEQ_ELTYPE_BITS bits contain the ID of the element type. The current version has CV_SEQ_ELTYPE_BITS equal to 5, that is, it supports up to 32 non-overlapping element types now. The file CVTypes.h declares the predefined types.
Example 14-6 Standard Types of Sequence Elements
#define CV_SEQ_ELTYPE_POINT #define CV_SEQ_ELTYPE_CODE #define CV_SEQ_ELTYPE_PPOINT #define CV_SEQ_ELTYPE_INDEX #define CV_SEQ_ELTYPE_GRAPH_EDGE &vtx_d */ #define CV_SEQ_ELTYPE_GRAPH_VERTEX #define CV_SEQ_ELTYPE_TRIAN_ATR */ #define CV_SEQ_ELTYPE_CONNECTED_COMP #define CV_SEQ_ELTYPE_POINT3D 1 2 3 4 5 /* /* /* /* /* (x,y) */ freeman code: 0..7 */ &(x,y) */ #(x,y) */ &next_o,&next_d,&vtx_o,

6 /* first_edge, &(x,y) / 7 / vertex of the binary tree 8 /* connected component 9 /* (x,y,z) / /

The next CV_SEQ_KIND_BITS bits, also 5 in number, specify the kind of the sequence. Again, predefined kinds of sequences are declared in the file CVTypes.h.
Example 14-7 Standard Kinds of Sequences
#define #define #define #define CV_SEQ_KIND_SET CV_SEQ_KIND_CURVE CV_SEQ_KIND_BIN_TREE CV_SEQ_KIND_GRAPH (0 (1 (2 (3 << << << << CV_SEQ_ELTYPE_BITS) CV_SEQ_ELTYPE_BITS) CV_SEQ_ELTYPE_BITS) CV_SEQ_ELTYPE_BITS)

14-26

OpenCV Reference Manual

Basic Structures and Operations Reference

The remaining bits are used to identify different features specific to certain sequence kinds and element types. For example, curves made of points (CV_SEQ_KIND_CURVE|CV_SEQ_ELTYPE_POINT), together with the flag CV_SEQ_FLAG_CLOSED belong to the type CV_SEQ_POLYGON or, if other flags are used, its subtype. Many contour processing functions check the type of the input sequence and report an error if they do not support this type. The file CVTypes.h stores the complete list of all supported predefined sequence types and helper macros designed to get the sequence type of other properties. Below follows the definition of the building block of sequences.
Example 14-8 CvSeqBlock Structure Definition
typedef struct CvSeqBlock { struct CvSeqBlock* prev; /* previous sequence block */ struct CvSeqBlock* next; /* next sequence block */ int start_index; /* index of the first element in the block + sequence->first->start_index */ int count; /* number of elements in the block */ char* data; /* pointer to the first element of the block */ } CvSeqBlock;

Sequence blocks make up a circular double-linked list, so the pointers prev and next are never NULL and point to the previous and the next sequence blocks within the sequence. It means that next of the last block is the first block and prev of the first block is the last block. The fields start_index and count help to track the block location within the sequence. For example, if the sequence consists of 10 elements and splits into three blocks of 3, 5, and 2 elements, and the first block has the parameter start_index = 2, then pairs <start_index, count> for the sequence blocks are <2,3>, <5,5>, and <10,2> correspondingly. The parameter start_index of the first block is usually 0 unless some elements have been inserted at the beginning of the sequence.

14-27

OpenCV Reference Manual

Basic Structures and Operations Reference

CreateSeq
Creates sequence.
CvSeq* cvCreateSeq(int seqFlags, int headerSize, int elemSize, CvMemStorage* storage); seqFlags

Flags of the created sequence. If the sequence is not passed to any function working with a specific type of sequences, the sequence value may be equal to 0, otherwise the appropriate type must be selected from the list of predefined sequence types. Size of the sequence header; must be more than or equal to sizeof(CvSeq). If a specific type or its extension is indicated, this type must fit the base type header. Size of the sequence elements in bytes. The size must be consistent with the sequence type. For example, for a sequence of points to be created, the element type CV_SEQ_ELTYPE_POINT should be specified and the parameter elemSize must be equal to sizeof(CvPoint). Sequence location.

headerSize

elemSize

storage

Discussion The function CreateSeq creates a sequence and returns the pointer to it. The function allocates the sequence header in the storage block as one continuous chunk and fills the parameter elemSize, flags headerSize, and storage with passed values, sets the parameter deltaElems (see the function SetSeqBlockSize) to the default value, and clears other fields, including the space behind sizeof(CvSeq).

NOTE. All headers in the memory storage, including sequence

headers and sequence block headers, are aligned with the 4-byte boundary.

14-28

OpenCV Reference Manual

Basic Structures and Operations Reference

SetSeqBlockSize
Sets up sequence block size.
void cvSetSeqBlockSize( CvSeq* seq, int blockSize ); seq blockSize

Sequence. Desirable block size.

Discussion The function SetSeqBlockSize affects the memory allocation granularity. When the free space in the internal sequence buffers has run out, the function allocates blockSize bytes in the storage. If this block immediately follows the one previously allocated, the two blocks are concatenated, otherwise, a new sequence block is created. Therefore, the bigger the parameter, the lower the sequence fragmentation probability, but the more space in the storage is wasted. When the sequence is created, the parameter blockSize is set to the default value ~1K. The function can be called any time after the sequence is created and affects future allocations. The final block size can be different from the one desired, e.g., if it is larger than the storage block size, or smaller than the sequence header size plus the sequence element size. The next four functions SeqPush,SeqPop,SeqPushFront,SeqPopFront add or remove elements to/from one of the sequence ends. Their time complexity is O(1), that is, all these operations do not shift existing sequence elements.

SeqPush
Adds element to sequence end.
void cvSeqPush( CvSeq* seq, void* element ); seq element

Sequence. Added element.

14-29

OpenCV Reference Manual

Basic Structures and Operations Reference

Discussion The function SeqPush adds an element to the end of the sequence. Although this function can be used to create a sequence element by element, there is a faster method (refer to Writing and Reading Sequences).

SeqPop
Removes element from sequence end.
void cvSeqPop( CvSeq* seq, void* element ); seq element

Sequence. Optional parameter. If the pointer is not zero, the function copies the removed element to this location.

Discussion The function SeqPop removes an element from the sequence. The function reports an error if the sequence is already empty.

SeqPushFront
Adds element to sequence beginning.
void cvSeqPushFront( CvSeq* seq, void* element ); seq element

Sequence. Added element.

Discussion The function SeqPushFront adds an element to the beginning of the sequence.

14-30

OpenCV Reference Manual

Basic Structures and Operations Reference

SeqPopFront
Removes element from sequence beginning.
void cvSeqPopFront( CvSeq* seq, void* element ); seq element

Sequence. Optional parameter. If the pointer is not zero, the function copies the removed element to this location.

Discussion The function SeqPopFront removes an element from the beginning of the sequence. The function reports an error if the sequence is already empty. Next two functions SeqPushMulti,SeqPopMulti are batch versions of the PUSH/POP operations.

SeqPushMulti
Pushes several elements to sequence end.
void cvSeqPushMulti(CvSeq* seq, void* elements, int count ); seq elements count

Sequence. Added elements. Number of elements to push.

Discussion The function SeqPushMulti adds several elements to the end of the sequence. The elements are added to the sequence in the same order as they are arranged in the input array but they can fall into different sequence blocks.

14-31

OpenCV Reference Manual

Basic Structures and Operations Reference

SeqPopMulti
Removes several elements from sequence end.
void cvSeqPopMulti( CvSeq* seq, void* elements, int count ); seq elements count

Sequence. Removed elements. Number of elements to pop.

Discussion The function SeqPopMulti removes several elements from the end of the sequence. If the number of the elements to be removed exceeds the total number of elements in the sequence, the function removes as many elements as possible.

SeqInsert
Inserts element in sequence middle.
void cvSeqInsert( CvSeq* seq, int beforeIndex, void* element ); seq beforeIndex

Sequence. Index before which the element is inserted. Inserting before 0 is equal to cvSeqPushFront and inserting before seq->total is equal to cvSeqPush. The index values in these two examples are boundaries for allowed parameter values. Inserted element.

element

Discussion The function SeqInsert shifts the sequence elements from the inserted position to the nearest end of the sequence before it copies an element there, therefore, the algorithm time complexity is O(n/2).

14-32

OpenCV Reference Manual

Basic Structures and Operations Reference

SeqRemove
Removes element from sequence middle.
void cvSeqRemove( CvSeq* seq, int index ); seq index

Sequence. Index of removed element.

Discussion The function SeqRemove removes elements with the given index. If the index is negative or greater than the total number of elements less 1, the function reports an error. An attempt to remove an element from an empty sequence is a specific case of this situation. The function removes an element by shifting the sequence elements from the nearest end of the sequence index.

ClearSeq
Clears sequence.
void cvClearSeq( CvSeq* seq );
seq

Sequence.

Discussion The function ClearSeq empties the sequence. The function does not return the memory to the storage, but this memory is used again when new elements are added to the sequence. This function time complexity is O(1).

14-33

OpenCV Reference Manual

Basic Structures and Operations Reference

GetSeqElem
Returns n-th element of sequence.
char* cvGetSeqElem( CvSeq* seq, int index, CvSeqBlock** block=0 ); seq index block

Sequence. Index of element. Optional argument. If the pointer is not NULL, the address of the sequence block that contains the element is stored in this location.

Discussion The function GetSeqElem finds the element with the given index in the sequence and returns the pointer to it. In addition, the function can return the pointer to the sequence block that contains the element. If the element is not found, the function returns 0. The function supports negative indices, where -1 stands for the last sequence element, -2 stands for the one before last, etc. If the sequence is most likely to consist of a single sequence block or the desired element is likely to be located in the first block, then the macro CV_GET_SEQ_ELEM (elemType, seq, index) should be used, where the parameter elemType is the type of sequence elements (CvPoint for example), the parameter seq is a sequence, and the parameter index is the index of the desired element. The macro checks first whether the desired element belongs to the first block of the sequence and, if so, returns the element, otherwise the macro calls the main function GetSeqElem. Negative indices always cause the cvGetSeqElem call.

SeqElemIdx
Returns index of concrete sequence element.
int cvSeqElemIdx( CvSeq* seq, void* element, CvSeqBlock** block=0 ); seq element

Sequence. Pointer to the element within the sequence.

14-34

OpenCV Reference Manual

Basic Structures and Operations Reference

block

Optional argument. If the pointer is not NULL, the address of the sequence block that contains the element is stored in this location.

Discussion The function SeqElemIdx returns the index of a sequence element or a negative number if the element is not found.

CvtSeqToArray
Copies sequence to one continuous block of memory.
void* cvCvtSeqToArray( CvSeq* seq, void* array, CvSlice slice=CV_WHOLE_SEQ(seq) ); seq array

Sequence. Pointer to the destination array that must fit all the sequence elements. Start and end indices within the sequence so that the corresponding subsequence is copied.

slice

Discussion The function CvtSeqToArray copies the entire sequence or subsequence to the specified buffer and returns the pointer to the buffer.

MakeSeqHeaderForArray
Constructs sequence from array.
void cvMakeSeqHeaderForArray( int seqType, int headerSize, int elemSize, void* array, int total, CvSeq* sequence, CvSeqBlock* block );

14-35

OpenCV Reference Manual

Basic Structures and Operations Reference

seqType headerSize

Type of the created sequence. Size of the header of the sequence. Parameter sequence must point to the structure of that size or greater size. Size of the sequence element. Pointer to the array that makes up the sequence. Total number of elements in the sequence. The number of array elements must be equal to the value of this parameter. Pointer to the local variable that is used as the sequence header. Pointer to the local variable that is the header of the single sequence block.

elemSize array total

sequence block

Discussion The function MakeSeqHeaderForArray, the exact opposite of the function CvtSeqToArray, builds a sequence from an array. The sequence always consists of a single sequence block, and the total number of elements may not be greater than the value of the parameter total, though the user may remove elements from the sequence, then add other elements to it with the above restriction.

Writing and Reading Sequences Reference

StartAppendToSeq
Initializes process of writing to sequence.
void cvStartAppendToSeq( CvSeq* seq, CvSeqWriter* writer ); seq writer

Pointer to the sequence. Pointer to the working structure that contains the current status of the writing process.

14-36

OpenCV Reference Manual

Basic Structures and Operations Reference

Discussion The function StartAppendToSeq initializes the writer to write to the sequence. Written elements are added to the end of the sequence. Note that during the writing process other operations on the sequence may yield incorrect result or even corrupt the sequence (see Discussion of the function FlushSeqWriter).

StartWriteSeq
Creates new sequence and initializes writer for it.
void cvStartWriteSeq(int seqFlags, int headerSize, int elemSize, CvMemStorage* storage, CvSeqWriter* writer); seqFlags

Flags of the created sequence. If the sequence is not passed to any function working with a specific type of sequences, the sequence value may be equal to 0, otherwise the appropriate type must be selected from the list of predefined sequence types. Size of the sequence header. The parameter value may not be less than sizeof(CvSeq). If a certain type or extension is specified, it must fit the base type header. Size of the sequence elements in bytes; must be consistent with the sequence type. For example, if the sequence of points is created (element type CV_SEQ_ELTYPE_POINT), then the parameter elemSize must be equal to sizeof(CvPoint). Sequence location. Pointer to the writer status.

headerSize

elemSize

storage writer

Discussion The function StartWriteSeq is the exact sum of the functions CreateSeq and StartAppendToSeq.

14-37

OpenCV Reference Manual

Basic Structures and Operations Reference

EndWriteSeq
Finishes process of writing.
CvSeq* cvEndWriteSeq( CvSeqWriter* writer); writer

Pointer to the writer status.

Discussion The function EndWriteSeq finishes the writing process and returns the pointer to the resulting sequence. The function also truncates the last sequence block to return the whole of unfilled space to the memory storage. After that the user may read freely from the sequence and modify it.

FlushSeqWriter
Updates sequence headers using writer state.
void cvFlushSeqWriter( CvSeqWriter* writer); writer

Pointer to the writer status.

Discussion The function FlushSeqWriter is intended to enable the user to read sequence elements, whenever required, during the writing process, e.g., in order to check specific conditions. The function updates the sequence headers to make reading from the sequence possible. The writer is not closed, however, so that the writing process can be continued any time. Frequent flushes are not recommended, the function SeqPush is preferred.

14-38

OpenCV Reference Manual

Basic Structures and Operations Reference

StartReadSeq
Initializes process of sequential reading from sequence.
void cvStartReadSeq( CvSeq* seq, CvSeqReader* reader, int reverse=0 ); seq reader reverse

Sequence. Pointer to the reader status. Whenever the parameter value equals 0, the reading process is going in the forward direction, that is, from the beginning to the end, otherwise the reading process direction is reverse, from the end to the beginning.

Discussion The function StartReadSeq initializes the reader structure. After that all the sequence elements from the first down to the last one can be read by subsequent calls of the macro CV_READ_SEQ_ELEM (elem, reader) that is similar to CV_WRITE_SEQ_ELEM. The function puts the reading pointer to the last sequence element if the parameter reverse does not equal zero. After that the macro CV_REV_READ_SEQ_ELEM (elem, reader) can be used to get sequence elements from the last to the first. Both macros put the sequence element to elem and move the reading pointer forward (CV_READ_SEQ_ELEM) or backward (CV_REV_READ_SEQ_ELEM). A circular structure of sequence blocks is used for the reading process, that is, after the last element has been read by the macro CV_READ_SEQ_ELEM, the first element is read when the macro is called again. The same applies to CV_REV_READ_SEQ_ELEM. Neither function ends reading since the reading process does not modify the sequence, nor requires any temporary buffers. The reader field ptr points to the current element of the sequence that is to be read first.

14-39

OpenCV Reference Manual

Basic Structures and Operations Reference

GetSeqReaderPos
Returns index of element to read position.
int cvGetSeqReaderPos( CvSeqReader* reader ); reader

Pointer to the reader status.

Discussion The function GetSeqReaderPos returns the index of the element in which the reader is currently located.

SetSeqReaderPos
Moves read position to specified index.
void cvSetSeqReaderPos( CvSeqReader* reader, int index, int is_relative=0 ); reader index is_relative

Pointer to the reader status. Position where the reader must be moved. If the parameter value is not equal to zero, the index means an offset relative to the current position.

Discussion The function SetSeqReaderPos moves the read position to the absolute or relative position. This function allows for cycle character of the sequence.

14-40

OpenCV Reference Manual

Basic Structures and Operations Reference

Sets Reference
Sets Functions

CreateSet
Creates empty set.
CvSet* cvCreateSet( int setFlags, int headerSize, int elemSize, CvMemStorage* storage); setFlags headerSize elemSize storage

Type of the created set. Set header size; may not be less than sizeof(CvSeq). Set element size; may not be less than 8 bytes, must be divisible by 4. Future set location.

Discussion The function CreateSet creates an empty set with a specified header size and returns the pointer to the set. The function simply redirects the call to the function CreateSeq.

SetAdd
Adds element to set.
int cvSetAdd( CvSet* set, CvSet* elem, CvSet** insertedElem=0 ); set elem

Set. Optional input argument, inserted element. If not NULL, the function copies the data to the allocated cell omitting the first 4-byte field. Optional output argument; points to the allocated cell.

insertedElem

14-41

OpenCV Reference Manual

Basic Structures and Operations Reference

Discussion The function SetAdd allocates a new cell, optionally copies input element data to it, and returns the pointer and the index to the cell. The index value is taken from the second 4-byte field of the cell. In case the cell was previously deleted and a wrong index was specified, the function returns this wrong index. However, if the user works in the pointer mode, no problem occurs and the pointer stored at the parameter insertedElem may be used to get access to the added set element.

SetRemove
Removes element from set.
void cvSetRemove( CvSet* set, int index ); set index

Set. Index of the removed element.

Discussion The function SetRemove removes an element with a specified index from the set. The function is typically used when set elements are accessed by their indices. If pointers are used, the macro CV_REMOVE_SET_ELEM(set, index, elem), where elem is a pointer to the removed element and index is any non-negative value, may be used to remove the element. Alternative way to remove an element by its pointer is to calculate index of the element via the function SeqElemIdx after which the function SetRemove may be called, but this method is much slower than the macro.

GetSetElem
Finds set element by index.
CvSetElem* cvGetSetElem( CvSet* set, int index );

14-42

OpenCV Reference Manual

Basic Structures and Operations Reference

set index

Set. Index of the set element within a sequence.

Discussion The function GetSetElem finds a set element by index. The function returns the pointer to it or 0 if the index is invalid or the corresponding cell is free. The function supports negative indices through calling the function GetSeqElem.

NOTE. The user can check whether the element belongs to the set

with the help of the macro CV_IS_SET_ELEM_EXISTS(elem) once the pointer is set to a set element.

ClearSet
Clears set.
void cvClearSet( CvSet* set ); set

Cleared set.

Discussion The function ClearSet empties the set by calling the function ClearSeq and setting the pointer to the list of free cells. The function takes O(1) time.

14-43

OpenCV Reference Manual

Basic Structures and Operations Reference

Sets Data Structures

Example 14-9 CvSet Structure Definition
#define CV_SET_FIELDS() CV_SEQUENCE_FIELDS() CvMemBlock* free_elems; typedef struct CvSet { CV_SET_FIELDS() } CvSet; \ \

Example 14-10 CvSetElem Structure Definition

#define CV_SET_ELEM_FIELDS() int* aligned_ptr; typedef struct _CvSetElem { CV_SET_ELEM_FIELDS() } CvSetElem; \

14-44

OpenCV Reference Manual

Basic Structures and Operations Reference

while( edge ) { edge = CV_NEXT_GRAPH_EDGE( edge, vertex ); count++; }.

The macro CV_NEXT_GRAPH_EDGE( edge, vertex ) returns the next edge after the edge incident to the vertex. The function reports an error, if input vertices do not belong to the graph, that makes it safer than GraphVtxDegreeByPtr, but less efficient.

GraphVtxDegreeByPtr
Finds edge in graph.
int cvGraphVtxDegreeByPtr( CvGraph* graph, CvGraphVtx* vtx ); graph vtx

Graph. Pointer to the graph vertex.

Discussion The function GraphVtxDegreeByPtr counts the edges incident to the graph vertex, both incoming and outcoming, and returns the result. To count the edges, the following code is used:
CvGraphEdge* edge = vertex->first; int count = 0; while( edge ) { edge = CV_NEXT_GRAPH_EDGE( edge, vertex ); count++; }.

The macro CV_NEXT_GRAPH_EDGE( edge, vertex ) returns the next edge after the edge incident to the vertex. The function is more efficient than GraphVtxDegree but less safe, because it does not check whether the input vertices belong to the graph.

14-52

OpenCV Reference Manual

Basic Structures and Operations Reference

ClearGraph
Clears graph.
void cvClearGraph( CvGraph* graph ); graph

Graph.

Discussion The function ClearGraph removes all the vertices and edges from the graph. Similar to the function ClearSet, this function takes O(1) time.

GetGraphVtx
Finds graph vertex by index.
CvGraphVtx* cvGetGraphVtx( CvGraph* graph, int vtxIdx ); graph vtxIdx

Graph. Index of the vertex.

Discussion
The function GetGraphVtx finds the graph vertex by index and returns the pointer to it or, if not found, to a free cell at this index. Negative indices are supported.

GraphVtxIdx
Returns index of graph vertex.
int cvGraphVtxIdx( CvGraph* graph, CvGraphVtx* vtx );

14-53

OpenCV Reference Manual

Basic Structures and Operations Reference

graph vtx

Graph. Pointer to the graph vertex.

Discussion The function GraphVtxIdx returns the index of the graph vertex by setting pointers to it.

GraphEdgeIdx
Returns index of graph edge.
int cvGraphEdgeIdx( CvGraph* graph, CvGraphEdge* edge ); graph edge

Graph. Pointer to the graph edge.

Discussion The function GraphEdgeIdx returns the index of the graph edge by setting pointers to it.

Graphs Data Structures

.
Example 14-11 CvGraph Structure Definition
#define CV_GRAPH_FIELDS() CV_SET_FIELDS() CvSet* edges; typedef struct _CvGraph { CV_GRAPH_FIELDS() } CvGraph; \ \

14-54

OpenCV Reference Manual

Basic Structures and Operations Reference

In OOP terms, the graph structure is derived from the set of vertices and includes a set of edges. Besides, special data types exist for graph vertices and graph edges.
Example 14-12 Definitions of CvGraphEdge and CvGraphVtx Structures
#define CV_GRAPH_EDGE_FIELDS() \ struct _CvGraphEdge* next[2]; \ struct _CvGraphVertex* vtx[2]; #define CV_GRAPH_VERTEX_FIELDS() struct _CvGraphEdge* first; typedef struct _CvGraphEdge { CV_GRAPH_EDGE_FIELDS() } CvGraphEdge; typedef struct _CvGraphVertex { CV_GRAPH_VERTEX_FIELDS() } CvGraphVtx; \

14-55

OpenCV Reference Manual

Basic Structures and Operations Reference

Matrix Operations Reference

Example 14-13 CvMat Structure Definition
typedef struct CvMat { int rows; // number of rows int cols; // number of cols CvMatType type; // type of matrix int step; // not used union { float* fl; //pointer to the float data double* db; //pointer to double-precision data }data; }CvMat

Example 14-14 CvMatArray Structure Definition

typedef struct CvMatArray { int rows; //number of rows int cols; //number pf cols int type; // type of matrices int step; // not used int count; // number of matrices in aary union { float* fl; float* db; }data; // pointer to matrix array data }CvMatArray

Alloc
Allocates memory for matrix data.
void cvmAlloc (CvMat* mat mat);

Pointer to the matrix for which memory must be allocated.

14-56

OpenCV Reference Manual

Basic Structures and Operations Reference

Discussion The function Alloc allocates memory for the matrix data.

AllocArray
Allocates memory for matrix array data.
void cvmAllocArray (CvMatArray* matAr matAr);

Pointer to the matrix array for which memory must be allocated.

Discussion The function AllocArray allocates memory for the matrix array data.

Free
Frees memory allocated for matrix data.
void cvmFree (CvMat* mat mat);

Pointer to the matrix.

Discussion The function Free releases the memory allocated by the function Alloc.

FreeArray
Frees memory allocated for matrix array data.
void cvmFreeArray (CvMat* matAr);

14-57

OpenCV Reference Manual

Basic Structures and Operations Reference

matAr

Pointer to the matrix array.

Discussion The function FreeArray releases the memory allocated by the function AllocArray.

Add
Computes sum of two matrices.
void cvmAdd ( CvMat* A, CvMat* B, CvMat* C); A B C

Pointer to the first source matrix. Pointer to the second source matrix. Pointer to the destination matrix.

Discussion The function Add adds the matrix A to B and stores the result in C.
( C = A + B, C ij = A ij + B ij ) .

Sub
Computes difference of two matrices.
void cvmSub ( CvMat* A, CvMat* B, CvMat* C); A B C

Pointer to the first source matrix. Pointer to the second source matrix. Pointer to the destination matrix.

Discussion The function Sub subtracts the matrix B from the matrix A and stores the result in C.

14-58

OpenCV Reference Manual

Basic Structures and Operations Reference

( C = A B, C ij = A ij B ij ) .

Scale
Multiplies matrix by scalar value.
void cvmScale ( CvMat* A, CvMat* B, double alpha ); A B alpha

Pointer to the source matrix. Pointer to the destination matrix. Scale factor.

Discussion The function Scale multiplies every element of the matrix by a scalar value
B = A , B ij = A ij .

DotProduct
Calculates dot product of two vectors in Euclidian metrics.
double cvmDotProduct(CvMat* A, CvMat* B); A B

Pointer to the first source vector. Pointer to the second source vector.

Discussion The function DotProduct calculates and returns the Euclidean dot product of two vectors.

i,j

DP = A B =

A ij B ij .

14-59

OpenCV Reference Manual

Basic Structures and Operations Reference

CrossProduct
Calculates cross product of two 3D vectors.
void cvmCrossProduct( CvMat* A, CvMat* B, CvMat* C); A B C

Pointer to the first source vector. Pointer to the second source vector. Pointer to the destination vector.

Discussion The function CrossProduct calculates the cross product of two 3D vectors:
C = A B , ( C 1 = A 2 B 3 A 3 B 2, C 2 = A 3 B 1 A 1 B 3, C 3 = A 1 B 2 A 2 B 1 )

Mul
Multiplies matrices.
void cvmMul ( CvMat* A, CvMat* B, CvMat* C ); A B C

Pointer to the first source matrix. Pointer to the second source matrix. Pointer to the destination matrix

Discussion The function Mul multiplies SrcA by SrcB and stores the result in Dst.

C = AB , C ij =

A ik B kj .

14-60

OpenCV Reference Manual

Basic Structures and Operations Reference

MulTransposed
Calculates product of matrix and transposed matrix.
void cvmMulTransposed (CvMat* A, CvMat* B, int order); A B order

Pointer to the source matrix. Pointer to the destination matrix. Order of multipliers.

Discussion The function MulTransposed calculates the product of A and its transposition. The function evaluates
B = A A
T T

if order is non-zero,

B = AA

otherwise.

Transpose
Transposes matrix.
void cvmTranspose ( CvMat* A, CvMat* B ); A B

Pointer to the source matrix. Pointer to the destination matrix.

Discussion The function Transpose transposes A and stores result in B.

B = A
T

Bij = A ji .

14-61

OpenCV Reference Manual

Basic Structures and Operations Reference

Invert
Inverts matrix.
void cvmInvert ( CvMat* A, CvMat* B ); A B

Pointer to the source matrix. Pointer to the destination matrix.

Discussion The function Invert inverts A and stores the result in B.

B = A , AB = BA = I ,
1

Trace
Returns trace of matrix.
double cvmTrace ( CvMat* A); A

Pointer to the source matrix.

Discussion The function Trace returns the sum of diagonal elements of the matrix A.

Det
Returns determinant of matrix.
double cvmDet ( CvMat* A);

trA =

Aii .

14-62

OpenCV Reference Manual

Basic Structures and Operations Reference

Pointer to the source matrix.

Discussion The function Det returns the determinant of the matrix A.

Copy
Copies one matrix to another.
void cvmCopy ( CvMat* A, CvMat* B ); A B

Pointer to the source matrix. Pointer to the destination matrix.

Discussion The function Copy copies the matrix A to the matrix B.

B = A, B ij = A ij .

SetZero
Sets matrix to zero.
void cvmSetZero ( CvMat* A ); A

Pointer to the matrix to be set to zero.

Discussion The function SetZero sets the matrix to zero.

A = 0, A ij = 0 .

14-63

OpenCV Reference Manual

Basic Structures and Operations Reference

SetIdentity
Sets matrix to identity.
void cvmSetIdentity ( CvMat* A ); A

Pointer to the matrix to be set to identity.

Discussion The function SetIdentity sets the matrix to identity.

A = I , A ij = ij =

1, i = j 0, i j

Mahalonobis
Calculates Mahalonobis distance between vectors.
double cvmMahalonobis A B T ( CvMat* A, CvMat* B, CvMat* T);

Pointer to the first source vector. Pointer to the second source vector. Pointer to the inverse covariance matrix.

Discussion The function Mahalonobis calculates the weighted distance between two vectors and returns it:
i, j

dist =

T ij ( A i Bi ) ( A j B j ) .

14-64

OpenCV Reference Manual

Basic Structures and Operations Reference

SVD
Calculates singular value decomposition.
void cvmSVD A V D ( CvMat* A, CvMat* V, CvMat* D);

Pointer to the source matrix. Pointer to the matrix where the orthogonal matrix is saved. Pointer to the matrix where the diagonal matrix is saved.

Discussion The function SVD decomposes the source matrix to product of two orthogonal and one diagonal matrices. , where U is an orthogonal matrix stored in A, D is a diagonal matrix, and V is another orthogonal matrix. If A is a square matrix, U and V are the same.
A = U DV
T

NOTE. The function SVD destroys the source matrix A. Therefore, in case the source matrix is needed after decomposition, clone it before running this function.

EigenVV
Computes eigenvalues and eigenvectors of symmetric matrix.
void cvmEigenVV ( CvMat* Src, CvMat* evects, CvMat* evals, Double eps); Src evects evals

Pointer to the source matrix. Pointer to the matrix where eigenvectors must be stored. Pointer to the matrix where eigenvalues must be stored.

14-65

OpenCV Reference Manual

Basic Structures and Operations Reference

eps

Accuracy of diagonalization.

Discussion The function EigenVV computes the eigenvalues and eigenvectors of the matrix Src and stores them in the parameters evals and evects correspondingly. Jacobi method is used. Eigenvectors are stored in successive rows of matrix eigenvectors. The resultant eigenvalues are in descending order.

NOTE. The function EigenVV destroys the source matrix Src.

Therefore, if the source matrix is needed after eigenvalues have been calculated, clone it before running the function EigenVV.

PerspectiveProject
Implements general transform of 3D vector array.
void cvmPerspectiveProject (CvMat* A, CvMatArray src, CvMatArray dst); A src dst

4x4 matrix. Source array of 3D vectors. Destination array of 3D vectors.

Discussion The function PerspectiveProject maps every input 3D vector T ( x' w, y' w, z' w ) , where and
w =

( x, y, z )

( x', y', z', w' ) = mat ( x, y , z , l )

w', w' 0

1, w' = 0

14-66

OpenCV Reference Manual

Basic Structures and Operations Reference

Drawing Primitives Reference

Line
Draws simple or thick line segment.
void cvLine( IplImage* img, CvPoint pt1, CvPoint pt2, int color, int thickness=1 ); img pt1 pt2 color thickness

Image. First point of the line segment. Second point of the line segment. Line color (RGB) or brightness (grayscale image). Line thickness.

Discussion The function Line draws the line segment between pt1 and pt2 points in the image. The line is clipped by the image or ROI rectangle. The Bresenham algorithm is used for simple line segments. Thick lines are drawn with rounding endings. To specify the line color, the user may use the macro CV_RGB (r, g, b) that makes a 32-bit color value from the color components.

LineAA
Draws antialiased line segment.
void cvLineAA( IplImage* img, CvPoint pt1, CvPoint pt2, int color, int scale=0 ); img pt1

Image. First point of the line segment.

14-67

OpenCV Reference Manual

Basic Structures and Operations Reference

pt2 color scale

Second point of the line segment. Line color (RGB) or brightness (grayscale image). Number of fractional bits in the end point coordinates.

Array of pointers to polygons. Array of polygon vertex counters. Number of contours that bind the filled region. Polygon color (RGB) or brightness (grayscale image).

Discussion The function FillPoly fills an area bounded by several polygonal contours. The function fills complex areas, for example, areas with holes, contour self-intersection, etc.

FillConvexPoly
Fills convex polygon.
void cvFillConvexPoly( IplImage* img, CvPoint* pts, int npts, int color ); img pts npts color

Image. Array of pointers to a single polygon. Polygon vertex counter. Polygon color (RGB) or brightness (grayscale image).

Discussion The function FillConvexPoly fills convex polygon interior. This function is much faster than the function FillPoly and fills not only the convex polygon but any monotonic polygon, that is, a polygon whose contour intersects every horizontal line (scan line) twice at the most.

14-72

OpenCV Reference Manual

Basic Structures and Operations Reference

PolyLine
Draws simple or thick polygons.
void cvPolyLine( IplImage* img, CvPoint** pts, int* npts, int contours, isClosed, int color, int thickness=1 ); img pts npts contours isClosed

Image. Array of pointers to polylines. Array of polyline vertex counters. Number of polyline contours. Indicates whether the polylines must be drawn closed. If closed, the function draws the line from the last vertex of every contour to the first vertex. Polygon color (RGB) or brightness (grayscale image). Thickness of the polyline edges.

color thickness

Discussion The function PolyLine draws a set of simple or thick polylines.

PolyLineAA
Draws antialiased polygons.
void cvPolyLineAA( IplImage* img, CvPoint** pts, int* npts, int contours, isClosed, int color, int scale=0 ); img pts npts contours

Image. Array of pointers to polylines. Array of polyline vertex counters. Number of polyline contours.

14-73

OpenCV Reference Manual

Basic Structures and Operations Reference

isClosed

Indicates whether the polylines must be drawn closed. If closed, the function draws the line from the last vertex of every contour to the first vertex. Polygon color (RGB) or brightness (grayscale image). Specifies number of fractional bits in the coordinates of polyline vertices.

color scale

Discussion The function PolyLineAA draws a set of antialiased polylines.

InitFont
Initializes font structure.
void cvInitFont( CvFont* font, CvFontFace fontFace, float hscale, float vscale, float italicScale, int thickness ); font fontFace

Pointer to the resultant font structure. Font name identifier. Only the font CV_FONT_VECTOR0 is currently supported. Horizontal scale. If equal to 1.0f, the characters have the original width depending on the font type. If equal to 0.5f, the characters are of half the original width. Vertical scale. If equal to 1.0f, the characters have the original height depending on the font type. If equal to 0.5f, the characters are of half the original height. Approximate tangent of the character slope relative to the vertical line. Zero value means a non-italic font, 1.0f means ~45 slope, etc. Thickness of lines composing letters outlines. The function cvLine is used for drawing letters.

hscale

vscale

italicScale

thickness

14-74

OpenCV Reference Manual

Basic Structures and Operations Reference

Discussion The function InitFont initializes the font structure that can be passed further into text drawing functions. Although only one font is supported, it is possible to get different font flavors by varying the scale parameters, slope, and thickness.

PutText
Draws text string.
void cvPutText( IplImage* img, const char* text, CvPoint org, CvFont* font, int color ); img text org font color

Input image. String to print. Coordinates of the bottom-left corner of the first letter. Pointer to the font structure. Text color (RGB) or brightness (grayscale image).

Discussion The function PutText renders the text in the image with the specified font and color. The printed text is clipped by ROI rectangle. Symbols that do not belong to the specified font are replaced with the rectangle symbol.

GetTextSize
Retrieves width and height of text string.
void cvGetTextSize( CvFont* font, const char* textString, CvSize* textSize, int* ymin ); font

Pointer to the font structure.

14-75

OpenCV Reference Manual

Basic Structures and Operations Reference

textString textSize

Input string. Resultant size of the text string. Height of the text does not include the height of character parts that are below the baseline. Lowest y coordinate of the text relative to the baseline. Negative, if the text includes such characters as g, j, p, q, y, etc., and zero otherwise.

ymin

Discussion The function GetTextSize calculates the binding rectangle for the given text string when a specified font is used.

Utility Reference

AbsDiff
Calculates absolute difference between two images.
void cvAbsDiff( IplImage* srcA, IplImage* srcB, IplImage* dst ); srcA srcB dst

First compared image. Second compared image. Destination image.

Discussion The function AbsDiff calculates absolute difference between two images.
dst ( x, y ) = abs ( srcA ( x, y ) srcB ( x, y ) ) .

14-76

OpenCV Reference Manual

Basic Structures and Operations Reference

AbsDiffS
Calculates absolute difference between image and scalar.
void cvAbsDiffS( IplImage* srcA, IplImage* dst, double value ); srcA dst value

Compared image. Destination image. Value to compare.

Discussion The function AbsDiffS calculates absolute difference between an image and a scalar.
dst (x,y) = abs ( srcA (x,y) value ) .

MatchTemplate
Fills characteristic image for given image and template.
void cvMatchTemplate( IplImage* img, IplImage* templ, IplImage* result, CvTemplMatchMethod method ); img templ

Image where the search is running. Searched template; must be not greater than the source image. The parameters img and templ must be single-channel images and have the same depth (IPL_DEPTH_8U, IPL_DEPTH_8S, or IPL_DEPTH_32F). Output characteristic image. It has to be a single-channel image with depth equal to IPL_DEPTH_32F. If the parameter img has the size of W H and the template has the size w h , the resulting image must have the size or selected ROI W w + 1 H h + 1 .

result

14-77

OpenCV Reference Manual

Basic Structures and Operations Reference

method

T ( x', y' ) I ( x + x', y + y' ) ,

T ( x', y' ) I ( x + x', y + y' )

T ( x ', y ' )

T ( x', y' )

I ( x + x', y + y' )

14-79

OpenCV Reference Manual

Basic Structures and Operations Reference

CvtPixToPlane
Divides pixel image into separate planes.
void cvCvtPixToPlane( IplImage* src, IplImage* dst0, IplImage* dst1, IplImage* dst2, IplImage* dst3); src dst0dst3

Source image. Destination planes.

Discussion The function CvtPixToPlane divides a color image into separate planes. Two modes are available for the operation. Under the first mode the parameters dst0, dst1, and dst2 are non-zero, while dst3 must be zero for the three-channel source image. For the four-channel source image all the destination image pointers are non-zero. In this case the function splits the three/four channel image into separate planes and writes them to destination images. Under the second mode only one of the destination images is not NULL; in this case, the corresponding plane is extracted from the image and placed into destination image.

CvtPlaneToPix
Composes color image from separate planes.
void cvCvtPlaneToPix( IplImage* src0, IplImage* src1, IplImage* src2, IplImage* src3, IplImage* dst ); src0src3 dst

Source planes. Destination image.

14-80

OpenCV Reference Manual

Basic Structures and Operations Reference

Discussion The function CvtPlaneToPix composes a color image from separate planes. If the dst has three channels, then src0, src1, and src2 must be non-zero, otherwise dst must have four channels and all the source images must be non-zero.

ConvertScale
Converts one image to another with linear transformation.
void cvConvertScale( IplImage* src, IplImage* dst, double scale, double shift); src dst scale shift

Source image. Destination image. Scale factor. Value added to the scaled source image pixels.

Discussion The function ConvertScale applies linear transform to all pixels in the source image and puts the result into the destination image with appropriate type conversion. The following conversions are supported:
IPL_DEPTH_8U IPL_DEPTH_32F, IPL_DEPTH_8U IPL_DEPTH_16S, IPL_DEPTH_8S IPL_DEPTH_32F, IPL_DEPTH_8S IPL_DEPTH_16S, IPL_DEPTH_16S IPL_DEPTH_32F, IPL_DEPTH_32S IPL_DEPTH_32F.

Applying the following formula converts integer types to float:

14-81

OpenCV Reference Manual

Basic Structures and Operations Reference

dst(x,y) = (float)(src(x,y)*scale + shift),

while the following formula does the other conversions:

dst(x,y) = saturate(round(src(x,y)*scale + shift)),

where round function converts the floating-point number to the nearest integer number and saturate function performs as follows:

Destination depth is IPL_DEPTH_8U: saturate(x)

255 : x

= x < 0 ? 0 : x > 255 ?

Destination depth is IPL_DEPTH_8S: saturate(x)

127 ? 127 : x

= x < -128 ? 128 : x >

Destination depth is IPL_DEPTH_16S: saturate(x)

x > 32767 ? 32767 : x

= x < -32768 ? 32768 :

Destination depth is IPL_DEPTH_32S: saturate(x)

= x.

InitLineIterator
Initializes line iterator.
int cvInitLineIterator( IplImage* img, CvPoint pt1, CvPoint pt2, CvLineIterator* lineIterator); img pt1 pt2 lineIterator

Image. Starting the line point. Ending the line point. Pointer to the line iterator state structure.

Discussion The function InitLineIterator initializes the line iterator and returns the number of pixels between two end points. Both points must be inside the image. After the iterator has been initialized, all the points on the raster line that connects the two ending points may be retrieved by successive calls of CV_NEXT_LINE_POINT point. The points on the

14-82

OpenCV Reference Manual

Basic Structures and Operations Reference

line are calculated one by one using the 8-point connected Bresenham algorithm. See Example 14-15 for the method of drawing the line in the RGB image with the image pixels that belong to the line mixed with the given color using the XOR operation.
Example 14-15 Drawing Line Using XOR Operation
void put_xor_line( IplImage* img, CvPoint pt1, CvPoint pt2, int r, int g, int b ) { CvLineIterator iterator; int count = cvInitLineIterator( img, pt1, pt2, &iterator); for( int i = 0; i < count; i++ ){ iterator.ptr[0] ^= (uchar)b; iterator.ptr[1] ^= (uchar)g; iterator.ptr[2] ^= (uchar)r; CV_NEXT_LINE_POINT(iterator); } }

SampleLine
Reads raster line to buffer.
int cvSampleLine( IplImage* img, CvPoint pt1, CvPoint pt2, void* buffer ); img pt1 pt2 buffer

Image. Starting the line point. Ending the line point. Buffer to store the line points; must have enough size to store MAX(|pt2.x - pt1.x| + 1,|pt2.y - pt1.y|+1) points.

Discussion The function SampleLine implements a particular case of application of line iterators. The function reads all the image points lying on the line between pt1 and pt2, including the ending points, and stores them into the buffer.

14-83

OpenCV Reference Manual

Basic Structures and Operations Reference

GetRectSubPix
Retrieves raster rectangle from image with sub-pixel accuracy.
void cvGetRectSubPix( IplImage* src, IplImage* rect, CvPoint2D32f center ); src rect center

Source image. Extracted rectangle; must have odd width and height. Floating point coordinates of the rectangle center. The center must be inside the image.

Discussion The function GetRectSubPix extracts pixels from src, if the pixel coordinates meet the following conditions:
center.x (widthrect-1)/2 <= x <= center.x + (widthrect-1)/2; center.y -(heightrect-1)/2 <= y <= center.y +(heightrect-1)/2.

Since the center coordinates are not integer, bilinear interpolation is applied to get the values of pixels in non-integer locations. Although the rectangle center must be inside the image, the whole rectangle may be partially occluded. In this case, the pixel values are spread from the boundaries outside the image to approximate values of occluded pixels.

bFastArctan
Calculates fast arctangent approximation for arrays of abscissas and ordinates.
void cvbFastAcrtan( const float* y, const float* x, float* angle, int len ); y x

Array of ordinates. Array of abscissas.

14-84

OpenCV Reference Manual

Basic Structures and Operations Reference

angle len

Calculated angles of points (x[i],y[i]). Number of elements in the arrays.

Discussion The function bFastAcrtan calculates an approximate arctangent value, the angle of the point (x,y). The angle is in the range from 0 to 360. Accuracy is about 0.1. For point (0,0) the resultant angle is 0.

Sqrt
Calculates square root of single float.
float cvSqrt( float x ); x

Scalar argument.

Discussion The function Sqrt calculates square root of a single argument. The argument should be non-negative, otherwise the result is unpredictable. The relative error is less than 9e-6.

bSqrt
Calculates square root of array of floats.
void cvbSqrt( const float* x, float* y, int len ); x y len

Array of arguments. Resultant array. Number of elements in the arrays.

14-85

OpenCV Reference Manual

Basic Structures and Operations Reference

Discussion The function cvbSqrt calculates the square root of an array of floats. The arguments should be non-negative, otherwise the results are unpredictable. The relative error is less than 3e-7.

InvSqrt
Calculates inverse square root of single float.
float cvInvSqrt( float x ); x

Scalar argument.

Discussion The function InvSqrt calculates the inverse square root of a single float. The argument should be positive, otherwise the result is unpredictable. The relative error is less than 9e-6.

bInvSqrt
Calculates inverse square root of array of floats.
void cvbInvSqrt( const float* x, float* y, int len ); x y len

Array of arguments. Resultant array. Number of elements in the arrays.

Discussion The function bInvSqrt calculates the inverse square root of an array of floats. The arguments should be positive, otherwise the results are unpredictable. The relative error is less than 3e-7.

14-86

OpenCV Reference Manual

Basic Structures and Operations Reference

bReciprocal
Calculates inverse of array of floats.
void cvbReciprocal( const float* x, float* y, int len ); x y len

Array of arguments. Resultant array. Number of elements in the arrays.

Discussion The function bReciprocal calculates the inverse (1/x) of arguments. The arguments should be non-zero. The function gives a very precise result with the relative error less than 1e-7.

bCartToPolar
Calculates magnitude and angle for array of abscissas and ordinates.
void cvbCartToPolar( const float* y, const float* x, float* mag, float* angle, int len ); y x mag angle len

Array of ordinates. Array of abscissas. Calculated magnitudes of points (x[i],y[i]). Calculated angles of points (x[i],y[i]). Number of elements in the arrays.

14-87

OpenCV Reference Manual

Basic Structures and Operations Reference

Discussion The function bCartToPolar calculates the magnitude x [ i ] 2 + y [ i ] 2 and the angle arctan ( y [ i ] x [ i ] ) of each point (x[i],y[i]). The angle is measured in degrees and varies from 0 to 360. The function is a combination of the functions bFastArctan and bSqrt, so the accuracy is the same as in these functions. If pointers to the angle array or the magnitude array are NULL, the corresponding part is not calculated.

bFastExp
Calculates fast exponent approximation for array of floats.
void cvbFastExp( const float* x, double* exp_x, int len); x exp_x len

Array of arguments. Array of results. Number of elements in the arrays.

Discussion The function bFastExp calculates fast exponent approximation for each element of the input array. The maximal relative error is about 7e-6.

bFastLog
Calculates fast approximation of natural logarithm for array of doubles.
void cvbFastLog( const double* x, float* log_x, int len); x log_x

Array of arguments. Array of results.

14-88

OpenCV Reference Manual

Basic Structures and Operations Reference

len

Number of elements in the arrays.

Discussion The function bFastLog calculates fast logarithm approximation for each element of the input array. Maximal relative error is about 7e-6.

RandInit
Initializes state of random number generator.
void cvRandInit( CvRandState* state, float lower, float upper, int seed ); state lower upper seed

Pointer to the initialized random number generator state. Lower boundary of uniform distribution. Upper boundary of uniform distribution. Initial 32-bit value to start a random sequence.

Discussion The function RandInit initializes the state structure that is used for generating uniformly distributed numbers in the range [lower, upper). A multiply-with-carry generator is used.

bRand
Fills array with random numbers.
void cvbRand( CvRandState* state, float* x, int len ); state x len

Random number generator state. Destination array. Number of elements in the array.

14-89

OpenCV Reference Manual

Basic Structures and Operations Reference

Discussion The function bRand fills the array with random numbers and updates generator state.

FillImage
Fills image with constant value.
void cvFillImage( IplImage* img, double val ); img val

Filled image. Value to fill the image.

Discussion The function FillImage is equivalent to either iplSetFP or iplSet, depending on the pixel type, that is, floating-point or integer.

RandSetRange
Sets range of generated random numbers without reinitializing RNG state.
void cvRandSetRange( CvRandState* state, double lower, double upper ); state lower upper

State of random number generator (RNG). New lower bound of generated numbers. New upper bound of generated numbers.

Discussion The function RandSetRange changes the range of generated random numbers without reinitializing RNG state. For the current implementation of RNG the function is equivalent to the following code:

14-90

OpenCV Reference Manual

Basic Structures and Operations Reference

unsigned seed = state.seed; unsigned carry = state.carry; cvRandInit( &state, lower, upper, 0 ); state.seed = seed; state.carry = carry;

However, the function is preferable because of compatibility with the next versions of the library.

KMeans
Splits set of vectors into given number of clusters.
void cvKMeans ( int numClusters, CvVect32f* samples, int numSamples, int vecSize, CvTermCriteria termcrit, int* cluster ); numClusters samples numSamples vecSize termcrit cluster

Number of required clusters. Pointer to the array of input vectors. Number of input vectors. Size of every input vector. Criteria of iterative algorithm termination. Characteristic array of cluster numbers, corresponding to each input vector.

Discussion The function KMeans iteratively adjusts mean vectors of every cluster. Termination criteria must be used to stop the execution of the algorithm. At every iteration the convergence value is computed as follows:
K 2

old _ mean i new _ mean i

i=1

14-91

OpenCV Reference Manual

Basic Structures and Operations Reference E < Termcrit . epsilon .

The function terminates if

14-92

System Functions
This chapter describes system library functions.
Table 15-1 System Library Functions
Name Description Loads versions of functions that are optimized for a specific platform. Retrieves information about the library.

LoadPrimitives

GetLibraryInfo

LoadPrimitives
Loads optimized versions of functions for specific platform.
int cvLoadPrimitives (char* dllName, char* processorType); dllName

Name of dynamically linked library without postfix that contains the optimized versions of functions Postfix that specifies the platform type: W7 for Pentium 4 processor, A6 for Intel Pentium II processor, M6 for Intel Pentium II processor, NULL for auto detection of the platform type.

processorType

15-1

OpenCV Reference Manual

System Functions

Discussion The function LoadPrimitives loads the versions of functions that are optimized for a specific platform. The function is automatically called before the first call to the library function, if not called earlier.

GetLibraryInfo
Gets the library information string.
void cvGetLibraryInfo (char** version, int* loaded, char** dllName); version

Pointer to the string that will receive the build date information; can be NULL. Postfix that specifies the platform type: W7 for Pentium 4 processor, A6 for Intel Pentium III processor, M6 for Intel Pentium II processor, NULL for auto detection of the platform type.

loaded

dllName

Pointer to the full name of dynamically linked library without path, could be NULL.

Discussion The function GetLibraryInfo retrieves information about the library: the build date, the flag that indicates whether optimized DLLs have been loaded or not, and their names, if loaded.

15-2

OpenCV Reference Manual

System Functions

15-3

Bibliography

This bibliography provides a list of publications that might be useful to the Intel Computer Vision Library users. This list is not complete; it serves only as a starting point. [Borgefors86] Gunilla Borgefors. Distance Transformations in Digital Images. Computer Vision, Graphics and Image Processing 34, 344-371 (1986). G. Bradski and J. Davis. Motion Segmentation and Pose Recognition with Motion History Gradients. IEEE WACV'00, 2000. P. J. Burt, T. H. Hong, A. Rosenfeld. Segmentation and Estimation of Image Region Properties Through Cooperative Hierarchical Computation. IEEE Tran. On SMC, Vol. 11, N.12, 1981, pp. 802-809. J. Canny. A Computational Approach to Edge Detection, IEEE Trans. on Pattern Analysis and Machine Intelligence, 8(6), pp. 679-698 (1986). J. Davis and Bobick. The Representation and Recognition of Action Using Temporal Templates. MIT Media Lab Technical Report 402, 1997.

[Bradski00] [Burt81]

[Canny86]

[Davis97]

[DeMenthon92] Daniel F. DeMenthon and Larry S. Davis. Model-Based Object Pose in 25 Lines of Code. In Proceedings of ECCV '92, pp. 335-343, 1992. [Fitzgibbon95] Andrew W. Fitzgibbon, R.B.Fisher. A Buyers Guide to Conic Fitting. Proc.5th British Machine Vision Conference, Birmingham, pp. 513-522, 1995.

[Horn81]

Berthold K.P. Horn and Brian G. Schunck. Determining Optical Flow. Artificial Intelligence, 17, pp. 185-203, 1981.

16-1

OpenCV Reference Manual

Bibliography

[Hu62] [Jahne97] [Kass88]

M. Hu. Visual Pattern Recognition by Moment Invariants, IRE Transactions on Information Theory, 8:2, pp. 179-187, 1962. B. Jahne. Digital Image Processing. Springer, New York, 1997. M. Kass, A. Witkin, and D. Terzopoulos. Snakes: Active Contour Models, International Journal of Computer Vision, pp. 321-331, 1988. J.Matas, C.Galambos, J.Kittler. Progressive Probabilistic Hough Transform. British Machine Vision Conference, 1998. A. Rosenfeld and E. Johnston. Angle Detection on Digital Curves. IEEE Trans. Computers, 22:875-878, 1973. Y. Rubner. C. Tomasi, L.J. Guibas. Metrics for Distributions with Applications to Image Databases. Proceedings of the 1998 IEEE International Conference on Computer Vision, Bombay, India, January 1998, pp. 59-66.

[Matas98] [Rosenfeld73] [RubnerJan98]

[RubnerSept98] Y. Rubner. C. Tomasi, L.J. Guibas. The Earth Movers Distance as a Metric for Image Retrieval. Technical Report STAN-CS-TN-98-86, Department of Computer Science, Stanford University, September 1998. [RubnerOct98] Y. Rubner. C. Tomasi. Texture Metrics. Proceeding of the IEEE International Conference on Systems, Man, and Cybernetics, San-Diego, CA, October 1998, pp. 4601-4607. http://robotics.stanford.edu/~rubner/publications.html [Serra82] [Schiele00] J. Serra. Image Analysis and Mathematical Morphology. Academic Press, 1982.

Bernt Schiele and James L. Crowley. Recognition without Correspondence Using Multidimensional Receptive Field Histograms. In International Journal of Computer Vision 36 (1), pp. 31-50, January 2000.
S. Suzuki, K. Abe. Topological Structural Analysis of Digital Binary Images by Border Following. CVGIP, v.30, n.1. 1985, pp. 32-46. C.H. Teh, R.T. Chin. On the Detection of Dominant Points on Digital Curves. - IEEE Tr. PAMI, 1989, v.11, No.8, p. 859-872.

[Suzuki85] [Teh89]

16-2

OpenCV Reference Manual

Bibliography

[Trucco98] [Williams92]

Emanuele Trucco, Alessandro Verri. Introductory Techniques for 3-D Computer Vision. Prentice Hall, Inc., 1998. D. J. Williams and M. Shah. A Fast Algorithm for Active Contours and Curvature Estimation. CVGIP: Image Understanding, Vol. 55, No. 1, pp. 14-26, Jan., 1992. http://www.cs.ucf.edu/~vision/papers/shah/92/WIS92A.pdf. A.Y.Yuille, D.S.Cohen, and P.W.Hallinan. Feature Extraction from Faces Using Deformable Templates in CVPR, pp. 104-109, 1989. Zhengyou Zhang. Parameter Estimation Techniques: A Tutorial with Application to Conic Fitting, Image and Vision Computing Journal, 1996.

[Yuille89] [Zhang96]

16-3

OpenCV Reference Manual

Bibliography

16-4

Supported Image Attributes and Operation Modes

The table below specifies what combinations of input/output parameters are accepted by different OpenCV functions. Currently, the table describes only array-processing functions, that is, functions, taking on input, output or both the structures IplImage and CvMat. Functions, working with complex data structures, e.g., contour processing, computational geometry, etc. are not included yet. Format is coded in form depth , where depth is coded as number of bits{u|s|f}, u stands for "integer Unsigned", s stands for "integer Signed" and f stands for "Floating point". For example, 8u means 8-bit unsigned image or array, 32f means floating-point image or array. 8u-64f is a short form of 8u, 8s, 16s, 32s, 32f, 64f. If a function has several input/output arrays, they all must have the same type unless opposite is explicitly stated. Word same in Output Format column means that the output array must have the same format with input array[s]. Word inplace in Output Format column means that the function changes content of one of the input arrays and thus produces the output. Word n/a means that the function output is not an image and format information is not applicable. Mask parameter, if present, must have format 8u or 8s. The following table includes only the functions that have raster images or matrices on input and/or on output.

A-1

OpenCV Reference Manual

Supported Image Attributes and Operation Modes

Table A-1

Image Atributes and Operation Modes for Array-Processing Functions

Number of Channels 1-4 1-4 1, 3 1-3

Function

Input Format

Output Format same same inplace

AbsDiff AbsDiffS Acc

8u - 64f 8u - 64f src = 8u, 8s, 32f acc = 32f (same channels number as src)

AdaptiveThreshold Add AddS And AndS bCartToPolar bFastArctan bFastExp bFastLog bInvSqrt bRand bReciprocal bSqrt CalcAffineFlowPyrLK CalcBackProject CalcEigenObjects CalcGlobalOrientation CalcHist

8u, 8s, 32f 8u - 64f 8u - 64f 8u - 64f 8u - 64f 32f 32f 32f 64f 32f
none

1 1-4 1-4 1-4 1-4 1 1 1 1 1 1 1 1 1 1 1 1 1

same same same same same

32f 32f 64f 64f 32f 32f 32f 32f 32f

same as img

32f 32f img = 8u histogram, img = 8u, 8s, 32f img = 8u mhi = 32f, orient = 32f, mask = 8u img = 8u, 8s, 32f

eig = 32f 32f histogram

A-2

OpenCV Reference Manual

Supported Image Attributes and Operation Modes

Table A-1

Image Atributes and Operation Modes for Array-Processing Functions (continued)

Number of Channels 1 1 1 1 1 1 1 1-4 1, 3 1-4 1-4 1-4

Function

Input Format

Output Format

CalcMotionGradient CalcOpticalFlowBM CalcOpticalFlowHS CalcOpticalFlowLK CalcOpticalFlowPyrLK CamShift Canny Circle CircleAA Cmp CmpS ConvertScale

mhi = 32f 8u 8u 8u img = 8u 8u, 8s, 32f 8u 8u - 64f 8u 8u - 64f 8u - 64f 8u - 64f

orient = 32f, mask 32f 32f 32f 32f

n/a

8u
inplace inplace

8u 8u 8u - 64f, the
same channels number

Copy CornerEigenValsAndVecs CornerMinEigenVal CountNonZero CrossProduct CvtPixToPlane

8u - 64f 8u, 8s, 32f 8u, 8s, 32f 8u - 64f 32f, 64f 8u - 64f

1-4 1 1 1-4 1 input - 2, 3 or 4, output - 1 input - 1, output 2,3 or 4 1 1, 3, 4 1 1-4

same

32f 32f 64f

same

8u - 64f

CvtPlaneToPix

8u - 64f

Det Dilate DistTransform DotProduct

32f, 64f 8u, 32f 8u, 8s 8u - 64f

64f
same

32f 64f

A-3

OpenCV Reference Manual

Supported Image Attributes and Operation Modes

Table A-1

Image Atributes and Operation Modes for Array-Processing Functions (continued)

Number of Channels 1-4 1 1-4 1, 3 1, 3, 4 1-4 1-4 1 1 1 1-4 1 1 1

Function

Input Format

Output Format inplace same inplace inplace same inplace inplace n/a contour n/a same inplace same or 32f or 64f for 8u & 8s n/a

DrawContours EigenVV Ellipse EllipseAA Erode FillConvexPoly FillPoly FindChessBoardCornerGuesses FindContours FindCornerSubPix Flip FloodFill GetRectSubPix GoodFeaturesToTrack

contour, img = 8u - 64f 32f, 64f 8u - 64f 8u 8u, 32f 8u - 64f 8u - 64f 8u img = 8u, 8s img = 8u, 8s, 32f 8u - 64f 8u, 32f 8u, 8s, 32f, 64f img = 8u, 8s, 32f, eig = 32f, temp = 32f img = 8u img = 8u img = 8u img = 8u 32f, 64f 8u, 8s, 32f 8u - 64f 8u 32f, 64f 8u, 8s, 32f

HoughLines HoughLinesP HoughLinesSDiv ImgToObs_DCT Invert Laplace Line LineAA Mahalonobis MatchTemplate

1 1 1 1 1 1 1-4 1, 3 1 1

n/a n/a n/a n/a same

16s, 32f
inplace inplace same

32f

A-4

OpenCV Reference Manual

Supported Image Attributes and Operation Modes

Table A-1

Image Atributes and Operation Modes for Array-Processing Functions (continued)

Number of Channels 1 1 1-4 1-4 1 1-4 1-4 1, 3, 4 1-4 1, 3 1-3

Function

Input Format

Output Format same same

MatMulAdd MatMulAddEx Mean Mean_StdDev MeanShift MinMaxLoc Moments MorphologyEx MulAddS MultiplyAcc

32f, 64f 32f, 64f 8u - 64f 8u - 64f 8u, 8s, 32f 8u - 64f (coi!=0) 8u - 64f (coi!=0) 8u, 32f 8u - 64f src = 8u, 8s, 32f acc = 32f (same channels number as src)

64f 64f
n/a

CvPoint, 64f CvMoments

same same inplace

Norm

8u - 64f (coi!=0, if mask!=0) 8u - 64f 8u - 64f 32f, 64f 8u - 64f 8u 8u, 8s, 32f 8u - 64f 8u, 8s, 32f 8u 8u, 8s, 32f
none

1-4

64f

Or OrS PerspectiveProject PolyLine PolyLineAA PreCornerDetect PutText PyrDown PyrSegmentation PyrUp RandNext

1-4 1-4 1 1-4 1, 3 1 1-4 1, 3 1, 3 1, 3 1

same same same inplace inplace

32f
inplace same same same

32u

A-5

OpenCV Reference Manual

Supported Image Attributes and Operation Modes

Table A-1

Image Atributes and Operation Modes for Array-Processing Functions (continued)

Number of Channels 1-4 1, 3 1-3

Function

Input Format

Output Format inplace inplace

Rectangle RunningAvg

8u - 64f src = 8u, 8s, 32f acc = 32f (same channels number as src)

SampleLine SegmentMotion Set SetIdentity SetZero SnakeImage Sobel SquareAcc

8u - 64f 32f 8u - 64f 32f, 64f 8u - 64f img = 8u, 8s, 32f 8u, 8s, 32f src = 8u, 8s, 32f acc = 32f (same channels number as src)

1-4 1 1-4 1 1-4 1 1 1, 3 1-3

inplace

32f
inplace same inplace n/a

16s, 32f
inplace

StartFindContours Sub SubRs SubS Sum SVD Threshold Transpose UnDistort UnDistortOnce

img = 8u, 8s 8u - 64f 8u - 64f 8u - 64f 8u - 64f 32f, 64f 8u, 8s, 32f 8u - 64f 8u 8u

1 1-4 1-4 1-4 1-4 1 1 1-4 1, 3 1, 3

contour same same same

64f
same same same same same

A-6

OpenCV Reference Manual

Supported Image Attributes and Operation Modes

Table A-1

Image Atributes and Operation Modes for Array-Processing Functions (continued)

Number of Channels 1 1-4 1-4

Function

Input Format

Output Format

UpdateMotionHistory Xor XorS

mhi = 32f, silh = 8u, 8s 8u - 64f 8u - 64f

mhi = 32f
same same

A-7

OpenCV Reference Manual

Supported Image Attributes and Operation Modes

A-8

Glossary
arithmetic operation background An operation that adds, subtracts, multiplies, or squares the image pixel values. A set of motionless image pixels, that is, pixels that do not belong to any object moving in front of the camera. This definition can vary if considered in other techniques of object extraction. For example, if a depth map of the scene is obtained, background can be defined as parts of scene that are located far enough from the camera. A region, either a positive or negative, that results from applying the Laplacian to an image. See Laplacian pyramid. An iterative pyramid-linking algorithm implementing a combined segmentation and feature computation. The algorithm finds connected components without a preliminary threshold, that is, it works on a grayscale image. Continuously Adaptive Mean-SHIFT algorithm. It is a modification of MeanShift algorithm that can track an object varying in size, e.g., because distance between the object and the camera varies. channel in the image to process. See channel of interest. A number of pixels sharing a side (or, in some cases, a corner as well). An area where level curves multiplied by the gradient magnitude assume a local maximum.

blob Burts algorithm

CamShift

channel of interest COI connected component corner

Glossary-1

OpenCV Reference Manual

Glossary

down-sampling

Down-sampling conceptually decreases image size by integer through replacing a pixel block with a single pixel. For instance, down-sampling by factor of 2 replaces a 2 X 2 block with a single pixel. In image processing convolution of the original image with blurring or Gaussian kernel precedes down-sampling. minimal work needed to translate one point mass configuration to another, normalized by the total configuration mass. The EMD is a optimal solution of transportation problem. A point at which the gradient assumes a local maximum along the gradient direction. See earth mover distance. Flood filling means that a group of connected pixels with close values is filled with, or is set to, a certain value. The flood filling process starts with some point, called seed, that is specified by function caller and then it propagates until it reaches the image ROI boundary or cannot find any new pixels to fill due to a large difference in pixel values. A set of images derived from each other with combination of convolution with Gaussian kernel and down-sampling. See down-sampling and up-sampling. A discrete approximation of stochastic variable probability distribution. The variable can be both a scalar value and a vector. Histograms represent a simple statistical description of an object, e.g., an image. The object characteristics are measured during iterations through that object See edge, ridge, and blob. A set of images, which can be obtained by subtracting upsampled images from the original Gaussian Pyramid, that is, Li = Gi up-sample (Gi+1) or Li = Gi up-sample

earth mover distance

edge EMD flood filling

Gaussian pyramid

histogram

image features Laplacian pyramid

Glossary-2

OpenCV Reference Manual

Glossary

(down-sample (Gi)), where Li are images from Laplacian Pyramid and Gi are images from Gaussian Pyramid. See also down-sampling and up-sampling. locally minimum A triangle made of two boundary runs in hierarchical interceptive area triangle representation of contours, if the interceptive area of its base line is smaller than both its neighboring triangles areas. LMIAT See locally minimum interceptive area triangle. mathematical morphologyA set-theoretic method of image analysis first developed by Matheron and Serra. The two basic morphological operations are erosion (thinning) and dilation (thickening). All operations involve an image A (object of interest) and a kernel element B (structuring element). memory storage Storage that provides the space for storing dynamic data structures. A storage consists of a header and a double-linked list of memory blocks treated as a stack, that is, the storage header contains a pointer to the block not occupied entirely and an integer value, the number of free bytes in this block.

minimal enclosing circle A circle in a planar point set whose points are entirely located either inside or on the boundary of the circle. Minimal means that there is no enclosing circle of a smaller radius. MHI motion history image See motion history image. Motion history image (MHI) represents how the motion took place. Each MHI pixel has a value of timestamp corresponding to the latest motion in that pixel. Very early motions, which occured in the past beyond a certain time threshold set from the current moment, are cleared out. As the person or object moves, copying the most recent foreground silhouette as the highest values in the motion history image creates a layered history of the resulting motion. An apparent motion of image brightness.

optical flow

Glossary-3

OpenCV Reference Manual

Glossary

pixel value

An integer or float point value that defines brightness of the image point corresponding to this pixel. For instance, in the case of 8u format images, the pixel value is an integer number from 0 to 255. A part of the image or a certain color plane in the image, or both. Sort of a skeletonized high contrast object within an image. Ridges are found at points where the gradient is non-zero (or the gradient is above a small threshold). See region of interest. A resizable array of arbitrary type elements located in the memory storage. The sequence is discontinuous. Sequence data may be divided into several continuous blocks, called sequence blocks, that can be located in different memory blocks. Generalization of histograms under which characteristic values with rather fine quantization are gathered and only non-zero bins are dynamically stored. An energy-minimizing parametric closed curve guided by external forces. Marking the image regions coinciding with the given template according to a certain rule (minimum squared difference or maximum correlation between the region and template). Lower and upper levels of pixel values corresponding to certain conditions. See pixel value. Up-sampling conceptually increases image size through replacing a single pixel with a pixel block. For instance, up-sampling by factor of 2 replaces a single pixel with a 2 X 2 block. In image processing convolution of the original image with Gaussian kernel, multiplied by the squared up-sampling factor, follows up-sampling.

region of interest ridge

G
Gabor transform, 3-29 Gaussian window, 2-19 GDI draw functions, 7-15 geometric image formation, 6-10 Geometry convexity defects, 4-16 ellipse fitting, 4-14 fitting of conic, 4-14 line fitting, 4-15 weighted least squares, 4-16 Geometry Data Types, 11-25 cvConvexityDefect Structure Definition, 11-25 Geometry Functions CalcPGH, 11-23 CheckContourConvexity, 11-21 ContourConvexHull, 11-18 ContourConvexHullApprox, 11-20 ConvexHull, 11-17 ConvexHullApprox, 11-18 ConvexityDefects, 11-21 FitEllipse, 11-12 FitLine2D, 11-13 FitLine3D, 11-15 MinAreaRect, 11-22 MinEnclosingCircle, 11-24 Project3D, 11-16 Gesture Recognition algorithm, 6-16 homography matrix, 6-18 image mask, 6-16 probability density, 6-17 Gesture Recognition Functions CalcImageHomography, 13-23 CalcProbDensity, 13-24 CreateHandMask, 13-23

FindHandRegion, 13-21 FindHandRegionA, 13-22 MaxRect, 13-25 graph non-oriented, 7-13 oriented, 7-13 graphs, 7-11 grayscale image, 3-11, 3-15, 3-20, 3-24, 7-2, 7-15 Greens formula, 4-5

H
hardware and software requirements, 1-3 header, 7-4, 7-10 hierarchical representation of contours, 4-8 Histogram analyzing shapes, 3-26 bayesian-based object recognition, 3-26 content based retrieval, 3-26 definition, 3-25 histogram back-projection, 2-10 signature, 3-27 Histogram Data Types, 10-57 Histogram Functions CalcBackProject, 10-51 CalcBackProjectPatch, 10-52 CalcContrastHist, 10-55 CalcEMD, 10-54 CalcHist, 10-50 CompareHist, 10-48 CopyHist, 10-49 CreateHist, 10-41 GetHistValue_1D, 10-45 GetHistValue_2D, 10-45 GetHistValue_3D, 10-46 GetHistValue_nD, 10-46 GetMinMaxHistValue, 10-47 MakeHistHeaderForArray, 10-42 NormalizeHist, 10-47 QueryHistValue_1D, 10-43 QueryHistValue_2D, 10-43 QueryHistValue_3D, 10-44

Index-5

OpenCV Reference Manual

Index

QueryHistValue_nD, 10-44 ReleaseHist, 10-42 SetHistThresh, 10-50 ThreshHist, 10-48 HMM, See Embedded Hidden Markov Models homography, 6-2 homography matrix, 6-2, 6-18 Horn & Schunck Technique, 2-19 Lagrangian multiplier, 2-19 HT, See Hough Transform in Features Hu invariants, 3-15 Hu moments, 6-18

IPL, See Intel Image Processing Library

L
Lagrange multiplier, 4-15 least squares method, 4-15 lens distortion, 6-2 distortion coefficients radial, 6-4 tangenial, 6-4 line fitting, 4-15 LMIAT, See locally minimum interceptive area triangle Lucas & Kanade Technique, 2-19

I
Image Functions, 7-1 Image Functions Reference CopyImage, 14-14 CreateImage, 14-8 CreateImageData, 14-10 CreateImageHeader, 14-7 GetImageRawData, 14-12 InitImageHeader, 14-13 ReleaseImage, 14-9 ReleaseImageData, 14-10 ReleaseImageHeader, 14-9 SetImageCOI, 14-11 SetImageData, 14-11 SetImageROI, 14-12 Image Statistics Functions CountNonZero, 10-20 GetCentralMoment, 10-25 GetHuMoments, 10-27 GetNormalizedCentralMoment, 10-26 GetSpatialMoment, 10-25 Mean, 10-21 Mean_StdDev, 10-21 MinMaxLoc, 10-22 Moments, 10-24 Norm, 10-22 SumPixels, 10-20 Intel Image Processing Library, 1-1, 7-1

M
Mahalanobis distance, 6-18 manual organization, 1-4 mathematical morphology, 3-19 Matrix Operations, 7-15 Matrix Operations Data Types cvMat Structure Definition, 14-56 cvMatArray Structure Definition, 14-56 Matrix Operations Functions Add, 14-58 Alloc, 14-56 AllocArray, 14-57 Copy, 14-63 CrossProduct, 14-60 Det, 14-62 DotProduct, 14-59 Free, 14-57 FreeArray, 14-57 Invert, 14-62 Mahalonobis, 14-64 Mul, 14-60 MulTransposed, 14-61 PerspectiveProject, 14-66 Scale, 14-59 SetIdentity, 14-64 SetZero, 14-63 Sub, 14-58

Index-6

OpenCV Reference Manual

Index

SVD, 14-65 Trace, 14-62 Transpose, 14-61 mean location, 2-11 Mean Shift algorithm, 2-9 memory block, 7-4 memory storage, 7-4 M-estimators, 4-15 MHT, See multidimesional Hough transform in Features model plane, 6-2 moire, 6-8 Morphology angle resolution, 3-29 black hat, 3-23 CIE Lab model, 3-29 closing equation, 3-21 dilation, 3-19 dilation formula, 3-20 dilation formula in 3D, 3-22 dilation in 3D, 3-21 Earth mover distance, 3-27 erision in 3D, 3-21 erosion, 3-19 erosion formula, 3-20 erosion formula in 3D, 3-23 flow matrix, 3-28 ground distance, 3-29 lower boundary of EMD, 3-30 morphological gradient function, 3-23 object of interest, 3-19 opening equation, 3-21 optimal flow, 3-28 scale resolution, 3-29 structuring element, 3-19 thickening, See dilation thinning, See erosion top hat, 3-23 Morphology Functions CreateStructuringElementEx, 10-30 Dilate, 10-32 Erode, 10-31

MorphologyEx, 10-33 ReleaseStructuringElement, 10-31 Motion History Image, 2-3 motion representation, 2-2 motion gradient image, 2-3 regional orientation, 2-6 motion segmentation, 2-7 downward stepping floodfill, 2-7 Motion Templates motion template images, 2-2 normal optical flow method, 2-2 Motion Templates Functions, 9-6 CalcGlobalOrientation, 9-7 CalcMotionGradient, 9-6 SegmentMotion, 9-8 UpdateMotionHistory, 9-6

N
node child, 4-10 parent, 4-10 root, 4-10 trivial, 4-13 node distance, 4-13 node weight, 4-13 non-coplanar points, See also non-degenerate points, 6-14 non-degenerate points, See also non-coplanar points, 6-14 non-maxima suppression, 4-3 notational conventions, 1-8

O
object model pseudoinverse, 6-14 online version, 1-8 optical flow, 2-18 Optical Flow Functions, 9-12 CalcOpticalFlowBM, 9-13 CalcOpticalFlowHS, 9-12

Index-7

OpenCV Reference Manual

Index

CalcOpticalFlowLK, 9-13 CalcOpticalFlowPyrLK, 9-14

P
parent node, 4-10 perspective distortion, 6-13 perspective model, 6-10 pinhole model, See perspective model Pixel Access Macros, 14-14 Pixel Access Macros Reference CV_INIT_PIXEL_POS, 14-16 CV_MOVE, 14-17 CV_MOVE_PARAM, 14-18 CV_MOVE_PARAM_WRAP, 14-18 CV_MOVE_TO, 14-16 CV_MOVE_WRAP, 14-17 Pixel Access Macros Structures cvPixelPosition Structures, 14-14 platforms supported, 1-4 polygonal approximation, 4-1 k-cosine curvature, 4-2 L1 curvature, 4-2 Rosenfeld-Johnston algorithm, 4-2 Teh and Chin algorithm, 4-3 POS, See pose from orthography ans scaling pose, 6-10 pose approximation method, 6-11 pose from orthography and scaling, 6-11 POSIT algorithm, 6-13 focal length, 6-14 geometric image formation, 6-10 object image, 6-14 object model, 6-14 pose approximation method, 6-11 pose from orthography and scaling, 6-11 POSIT algorithm, 6-9 POSIT Functions CreatePOSITObject, 13-19 POSIT, 13-19

ReleasePOSITObject, 13-20 PPHT, See progressive probabilistic Hough transform in Features prefix, in function names, 1-9, 1-10 PUSH version, 7-6 Pyramid, 10-56 Pyramid Data Types cvConnectedComp Structure Definition, 10-56 Pyramid Functions PyrDown, 10-28 PyrSegmentation, 10-29 PyrUp, 10-28 Pyramids down-sampling, 3-15 Gaussian, 3-15 image segmentation, 3-17 hierarchical computing structure, 3-17 hierarchical smoothing, 3-17 segmentation, 3-17 Laplacian, 3-15 son-father relationship, 3-17 up-sampling, 3-16

R
radial distortion, 6-2 radial distortion coefficients, 6-4 region of interest, 7-3 related publications, 1-8 RLE coding, 4-1 ROI, See region of interest root node, 4-10 Rosenfeld-Johnston algorithm, 4-2 rotation matrix, 6-5 rotation vector, 6-5

S
scalar factor, 6-3 scaled orthographic projection, See also weak-perspective projection model, 6-11

Index-8

OpenCV Reference Manual

Index

scanlines, 6-6 Sequence Reference, 14-25 sequences, 7-5 sets, 7-8 shape partitioning, 4-12 SHT, See standard Hough transform in Features stochastic variable, 3-15 synthesized image, 6-5 System Functions GetLibraryInfo, 15-2 LoadPrimitives, 15-1

CvtPlaneToPix, 14-80 FillImage, 14-90 GetRectSubPix, 14-84 InitLineIterator, 14-82 InvSqrt, 14-86 KMeans, 14-91 MatchTemplate, 14-77 RandInit, 14-89 RandSetRange, 14-90 SampleLine, 14-83 Sqrt, 14-85

V T
tangential distortion coefficients, 6-4 Teh and Chin algorithm, 4-3 three sigmas rule, 2-1 Threshold Functions AdaptiveThreshold, 10-36 Threshold, 10-38 trivial node, 4-13 vectoring algorithms, 3-1 View Morphing, 6-5 moire, 6-8 scanlines, 6-6 warped image, 6-6 view morphing algorithm, 6-6 View Morphing Functions DeleteMoire, 13-18 DynamicCorrespondMulti, 13-15 FindFundamentalMatrix, 13-12 FindRuns, 13-14 MakeAlphaScanlines, 13-16 MakeScanlines, 13-13 MorphEpilinesMulti, 13-16 PostWarpImage, 13-17 PreWarpImage, 13-13

U
Use of Eigen Object Functions, 12-7 Use of Eigen Objects Functions cvCalcEigenObjects in Callback Mode, 12-9 cvCalcEigenObjects in Direct Access Mode, 12-8 Utility Functions AbsDiff, 14-76 AbsDiffS, 14-77 bCartToPolar, 14-87 bFastArctan, 14-84 bFastExp, 14-88 bFastLog, 14-88 bInvSqrt, 14-86 bRand, 14-89 bReciprocal, 14-87 bSqrt, 14-85 ConvertScale, 14-81 CvtPixToPlane, 14-80

W
warped image, 6-6 weak-perspective projection, 6-12 weak-perspective projection model, 6-11 weighted least squares, 4-16 world coordinate system, 6-2

Index-9