International Journal of Computer Science & Information Technology (IJCSIT), Vol 2, No 6, December 2010

SEGMENTATION AND OBJECT RECOGNITION

USING EDGE DETECTION TECHNIQUES

Y.Ramadevi, T.Sridevi, B.Poornima, B.Kalyani

Department of CSE , Chaitanya Bharathi Institute of Technology

Gandipet, Hyderabad.
yrd@cbit.ac.in

ABSTRACT
Image segmentation is to partition an image into meaningful regions with respect to a particular
application. Object recognition is the task of finding a given object in an image or video sequence. In this
paper, interaction between image segmentation (using different edge detection methods) and object
recognition are discussed. Edge detection methods such as Sobel, Prewitt, Roberts, Canny, Laplacian of
Guassian(LoG) are used for segmenting the image. Expectation-Maximization (EM) algorithm, OSTU
and Genetic algorithms were used to demonstrate the synergy between the segmented images and object
recognition.

KEYWORDS
EM algorithm, OSTU, Genetic Algorithm, Image Segmentation, Object Recognition.

1. INTRODUCTION
Image segmentation is the foundation of object recognition and computer vision. In general,
image noise should be eliminated through image preprocessing. And there is some specifically-
given work (such as region extraction and image marking) to do after the main operation of
image segmentation for the sake of getting better visual effect. Two major computer vision
problems, image segmentation and object recognition, have been traditionally dealt with using a
strict, bottom-up ordering.
Image segmentation is the process of partitioning/subdividing a digital image into multiple
meaningful regions or sets of pixels regions with respect to a particular application. The
segmentation is based on measurements taken from the image and might be grey level, colour,
texture, depth or motion. The result of image segmentation is a set of segments that collectively
cover the entire image. All the pixels in region are similar with respect to some characteristic or
computed property, such as color, intensity, or texture. Adjacent regions differ with respect to
same characteristics. Edge detection is one of the frequently used techniques in digital image
processing.
Object recognition is the task of finding a given object in an image or video sequence. For any
object in an image, there are many 'features' which are interesting points on the object that can
be extracted to provide a "feature" description of the object. This description extracted from a
training image can then be used to identify the object when attempting to locate the object in a
test image containing many other objects[3][4].
Image segmentation is done using various edge detection techniques such as Sobel, Prewitt,
Roberts, Canny, LoG, EM algorithm, OSTU and Genetic Algorithm [1][2]. Implementation
DOI : 10.5121/ijcsit.2010.2614 153
International Journal of Computer Science & Information Technology (IJCSIT), Vol 2, No 6, December 2010

between Segmentation and Object recognition using the said techniques is studied [1].
Experimentation is performed on gray scale image using MATLAB 7.9. In this paper, Section 2
discusses about segmentation, and comparison of various edge detection techniques. Section 3
discusses about image segmentation and object recognition using other techniques. The
implementation and results are shown in section 4. The paper is concluded in section 5.

2. SEGMENTATION
Segmentation is the process of partitioning a digital image into its constituent parts or objects or
regions. These regions share common characteristics based on color, intensity, texture, etc. The
first step in image analysis is to segment an image based on discontinuity detection technique
(Edge-based) or similarity detection technique (Region-based). In discontinuity detection
technique, one approach is to partition an image based on abrupt changes in intensity near the
edges and it is known as Edge-based segmentation. In Similarity detection technique, region-
based segmentation partitions an image into regions that are similar according to a set of
predefined criteria. Region-based segmentation looks for uniformity within a sub-region, based
on a desired property, e.g., intensity, color, and texture as shown figure 1. The difference
between Region-based segmentation and edge-based segmentation is shown in Table 1.
Active contours are popular technique for image segmentation. An advantage of active contours
as image segmentation methods is that they partition an image into sub-regions with continuous
boundaries. There are two kinds of active contour models: Edge-based active contours use an
edge detector, usually based on the image gradient, to find the boundaries of sub-regions and to
attract the contours to the detected boundaries. Region-based active contours use the statistical
information of image intensity within each subset instead of searching geometrical boundaries

Original Image Region Based Segmentation

Figure 1. Region based Segmentation

Table 1. Differences between Region-based segmentation and edge-based segmentation

Region-based segmentation Edge based segmentation
Closed boundaries Boundaries formed not necessarily closed
Multi-spectral images improve No significant improvement for multi-spectral
segmentation images
Computation based on similarity Computation based on difference
2.1. Edge Detection for Image Segmentation
An Edge in an image is a significant local change in the image intensity, usually associated with
a discontinuity in either the image intensity or the first derivative of the image intensity. The
three steps in Edge detection process is[6] a) Filtering b) Enhancement and c) Detection.

154
International Journal of Computer Science & Information Technology (IJCSIT), Vol 2, No 6, December 2010

a. Filtering: Images are corrupted by noise such as salt and pepper noise, impulse noise and
Gaussian noise. As there is a trade-off between edge strength and noise reduction, filtering
is done.
b. Enhancement: It emphasizes pixels where there is a significant change in local intensity
values and is usually performed by computing the gradient magnitude.
c. Detection: Many points in an image have a nonzero value for the gradient, and not all of
these points are edges for a particular application. Thresholding is used for the detection of
edge points.
The different edge detection methods used are Sobel, Prewitt, Roberts, Canny, LoG, EM
algorithm, OSTU algorithm and Genetic Algorithm.
2.1.1. Sobel Operator: It performs 2-D spatial gradient measurement on an image and so
emphasizes regions of high spatial frequency that correspond to edges. The convolution mask
of Sobel operator are as shown Figure 2, which are used to obtain the gradient magnitude of
the image from the original

1 2 1 -1 0 1
0 0 0 -2 0 2
-1 -2 -1 -1 0 1
Figure 2. Sobel Mask

2.1.2. Prewitt Operator: The prewitt operator is an approximate way to estimate the
magnitude and orientation of the edge. The convolution mask of prewitt operator is shown in
figure 3.

1 1 1 -1 0 1
0 0 0 -1 0 1
-1 -1 -1 -1 0 1
Figure 3. Prewitt Mask

2.1.3. Roberts Operator: It performs 2-D spatial gradient measurement on an image. It

highlights regions of high spatial frequency which often correspond to edges. The cross
convolution mask is shown in figure 4.

-1 0 0 1
0 -1 -1 0
Figure 4. Roberts Mask

2.1.4. Laplacian of Guassian(LoG) Operator: It is a second order derivative. The digital

implementation of the Laplacian function is made using the mask given in figure 5.

0 -1 0
-1 4 -1
0 -1 0
Figure 5. Laplacian of Guassian(LoG) Operator

155
International Journal of Computer Science & Information Technology (IJCSIT), Vol 2, No 6, December 2010

2.1.5. Canny Operator: It is a method to find edges by isolating noise from the image without
affecting the features of the edges in the image and then applying the tendency to find the
edges and the critical value for threshold.

3. IMAGE SEGMENTATION & OBJECT RECOGNITION

In Image Segmentation the generative models are used to decide which part of the image a
model should occupy. Active Appearance Model (AAMs) is used as generative models and
addresses the problem of jointly detecting and segmenting objects in images [1][5]. Regarding
recognition, each object hypothesis is validated based on the image area assigned to the object,
as well as the estimated model parameters, which indicate the familiarity of the object
appearance. On one hand, knowing the area occupied by an object is needed for the estimation
of the model parameters and, on the other hand, the model synthesis is used to assign
observations to the model. Since neither is known in advance, we cannot address each problem
separately. We view this problem as an instance of the broader problem of parameter estimation
with missing data: In our case, the missing data are the assignments of observations to models.
A well-known tool for addressing such problems is the EM algorithm.
An expectation-maximization (EM) algorithm is used for finding maximum likelihood
estimates of parameters in probabilistic models, where the model depends on unobserved latent
variables. In order to find maximum likelihood estimate we have to find probability density
function and loglikelihood.
3.1. The Expectation-Maximization Algorithm
The EM algorithm is an efficient iterative procedure to compute the Maximum Likelihood
Estimate (MLE) in the presence of missing or hidden data. Each iteration of the EM algorithm
consists of two processes: The E-step, and the M-step. In the expectation, or E-step, the missing
data are estimated given the observed data and current estimate of the model parameters. This is
achieved using the conditional expectation, explaining the choice of terminology. In the M-step,
the likelihood function is maximized under the assumption that the missing data are known. The
estimates of the missing data from the E-step are used in lieu of the actual missing data.
The EM algorithm seeks to find the MLE by iteratively applying the following two steps:
Expectation step: Calculate the expected value of the log likelihood function, with respect to
the conditional distribution of z given x under the current estimate of the parameters θ(t).
Maximization step: Find the parameter which maximizes this quantity.
3.2. Image segmentation using OTSU
Threshold selection is used in OTSU algorithm. The OTSU method is one of the applied
methods of image segmentation in selecting threshold automatically for its simple calculation
and good adaptation.
In image processing, OTSU’s thresholding method is used for automatic binarization level
decision, based on the shape of the histogram. The algorithm assumes that the image is
composed of two basic classes: Foreground and Background. It then computes an optimal
threshold value that minimizes the weighted within class variances of these two classes. It is

156
International Journal of Computer Science & Information Technology (IJCSIT), Vol 2, No 6, December 2010

mathematically proven that minimizing the within class variance is same as maximizing the
between class variance.
The different thresholding techniques are

1. Histogram shape-based methods.

2. Clustering-based methods.
3. Entropy-based methods.
4. Object attribute-based method.
5. Spatial methods.
6. Local methods.

The major problem with thresholding is that we consider only the intensity, not any
relationships between the pixels. There is no guarantee that the pixels identified by the
thresholding process are contiguous.

The steps of the OTSU algorithm: For each potential threshold T,

1. Separate the pixels into two clusters according to the threshold.
2. Find the mean of each cluster.
3. Square the difference between the means.
4. Multiply by the number of pixels in one cluster times the number in the other.

3.3 Image Segmentation using Genetic Algorithm

Genetic Algorithms (GAs) can be seen as a software tool that tries to find structure in data that
might seem random, or to make a seemingly unsolvable problem more or less 'solvable'. GAs
can be applied to domains about which there is insufficient knowledge or the size and/or
complexity is too high for analytic solution.
Basically, a genetic algorithm consists of three major operations: selection, crossover, and
mutation. The selection evaluates each individual and keeps only the fittest ones in the
population. In addition to those fittest individuals, some less fit ones could be selected according
to a small probability. The others are removed from the current population. The crossover
recombines two individuals to have new ones which might be better. The mutation operator
induces changes in a small number of chromosomes units. Its purpose is to maintain the
population diversified enough during the optimization process
The existing GA’s are founded upon the following main principles:
1. Reproduction
2. Fitness
3. Crossover
4. Mutation
The algorithms code crossover as either a swap of one bit, which is more like a single-point
mutation than a 'real crossover', or of several bits (used with genetic programming) and a
distinction is made between two parents (bit strings, but called chromosomes in GA
terminology) who are the identical, two different parents and single parent. Anyhow, the process
has the following procedure:

157
International Journal of Computer Science & Information Technology (IJCSIT), Vol 2, No 6, December 2010

1. Select two bit strings (chromosomes), or in case of the genetic programming: select a
branch of each parent.
2. Cut the chromosome (or branch) at a particular location.
3. Swap the bits/branches of the two parents.

4. IMPLEMENTATION AND RESULTS

Implementation of the techniques was done on different images. Colored images were converted
into gray scale image and then segmentation and recognition methods were applied. A sample
grey scale image (figure 6) is considered for segmentation and object recognition using Sobel,
Prewitt, Roberts, Canny, LoG , EM algorithm, OSTU algorithm and Genetic Algorithm. The
segmented image and recognized image using different edge operators’ techniques are shown in
Table 2.

Figure 6. Original Image

Table 2. Segmented and Recognized images using different operators/algorithms

Operator/ Algorithm Segmented Image Recognized Image

Sobel

Prewitt

158
International Journal of Computer Science & Information Technology (IJCSIT), Vol 2, No 6, December 2010

Roberts

Laplacian of Gaussian
(LoG)

Canny

EM Algorithm

159
International Journal of Computer Science & Information Technology (IJCSIT), Vol 2, No 6, December 2010

OSTU Algorithm

Genetic Algorithm

Canny operator performed better than Sobel, Prewitt, Roberts and LoG. The EM algorithm
produced stable segmentation effect on different types of images. OTSU showed good and
stable segmentation effect. Genetic algorithm exhibited normal segmentation effect on all types
of images.
5. CONCLUSIONS
This paper focuses mainly on the Image segmentation using edge operators. The interaction
between image segmentation and object recognition in the framework of the Sobel, Prewitt,
Roberts, Canny, LoG, Expectation-Maximization (EM) algorithm, OSTU Algorithm and
Genetic Algorithm are studied. MATLAB 7.9. was used for experimentation image.
Expectation-Maximization algorithm and OTSU algorithm exhibited stable segmentation effect.

6. REFERENCES
[1] Iasonas Kokkinos, and Petros Maragos (2009),”Synergy between Object Recognition and image
segmentation using Expectation and Maximization Algorithm”., IEEE Trans. on Pattern
Analysis and Machine Intelligence (PAMI), Vol. 31(8), pp. 1486-1501, 2009.
[2] Wen-Xiong Kang, Qing-Qiang Yang, Run-Peng Liang (2009), “The Comparative Research
on Image Segmentation Algorithms,” First International Workshop on Education Technology
and Computer Science.
[3] V. Ferrari, T. Tuytelaars, and L.V. Gool(2004), “Simultaneous Object Recognition and
Segmentation by Image Exploration,” Proc. Eighth European Conf. Computer Vision, 2004.
[4] B. Leibe, A. Leonardis, and B. Schiele(2004), “Combined Object Categorization and
Segmentation with an Implicit Shape Model,” Proc. ECCV Workshop Statistical Learning in
Computer Vision, 2004.
[5] Y.Ramadevi, B.Kalyani, T.Sridevi(2010), “ Synergy between Object Recognition and Image
Segmentation”, International Journal on Computer Science and Engineering, Vol. 02, No. 08,
2010, 2767-2772.
[6] N.Senthilkumarn, R.Rajesh(2009), ”Edge Detection Techniques for Image Segmentation- A
Survey of Soft Computing Approaches”, IJRTE, vol1,No2, 2009 250-254.

160
International Journal of Computer Science & Information Technology (IJCSIT), Vol 2, No 6, December 2010

Y Rama Devi received B.E. from Osmania University in 1991 and M.Tech (CSE)
degree from JNT University in 1997. She received his Ph.D. degree Central University,
Hyderabad in 2009. She is Professor, Chaitanya Bharathi Institute of Technology,
Hyderabad. Her research interests include Image Processing, Soft Computing, Data
Mining, and Bio-Informatics. She is a member for IEEE, ISTE, IETE, and IE. She has
published more than 25 research publications in various National, Inter-national
conferences, proceedings and Journals.

T.Sridevi received B E from Osmania University in 1992 and M.Tech (CSE) degree from
JNT University in 2002. She is pursuing her Ph D from Osmania University in Computer
Science under the guidance of Dr V. Vijaya Kumar. She has 14 years of teaching/industry
experience. She joined as Assistant Professor in Chaitanya Bharathi Institute of
Technology, Hyderabad, India in 2002. Presently she is an Associate Professor, Chaitanya
Bharathi Institute of Technology, Hyderabad. Her research areas include Water Marking,
Image Processing and Soft Computing. She is a life member of IETE. She has published
more than 5 research publications in various National, Inter-national conferences,
proceedings and Journals.

B.Poornima received her B.Tech from J.N.T.U in 2005. She is pursuing her M.Tech from
CBIT under Osmania University. Her area of interest is image processing.

B.Kalyani received B.Sc computers from Sri Krishnadevaraya University, Anantapur in

2003, M.Sc Mathematics from Sri Krishnadevaraya University, Anantapur in 2006 and
M.Tech(C.S.E) from Osmania University. Her area of interest is image processing.

161