Puter Vision

CS201 Lecture 02
Computer Vision:
Image Formation and Basic
Techniques
John Magee
1
Computer Vision
How are Computer Graphics and Computer Vision Related?
Recall: Computer graphics in general

Description of scene  Visual representation (Image)
Computer Vision in general:
Image(s)  Some description of the scene
Example -
Input: Image
Output: Face locations
Fujifilm camera demo 2

Data Structures for Images
 2D array vs. 1D array
 Interleaved RGB vs. Planar RGB
 Data stored in arrays vs. pointers to pixel
class/structure.
3
Some Easy Techniques
 Color Analysis
 Motion Analysis
 Template matching
(Some extra detail on the next few slides)
4
Color Analysis
Skin color analyzed by lookup of 2D histogram:
Histogram can be updated

during operation 
5
Motion Analysis
Motion analysis by frame differencing:
Recall: Video compression uses frame differencing. 6

Template Matching
Sum of Absolute Differences
 Normalized correlation coefficient matching over

multi-resolution search space.
12 x 16 Template
matching over all
resolutions 
7
Face Tracking
8
Face Detection vs. Face
Recognition
Face Detection exploits the similarities
between human faces.
- Using Probabilistic/Statistical Matching
Face Recognition exploits the differences
between human faces.
- Using Principle Component Analysis
9
Gaze Analysis
Eye (m x n) image
difference projected
Right Eye to x-axis:
Mirrored
Left Eye
Looking Looking
Left Straight
Computer Vision
What can go wrong?
– You might not know anything about a scene!

– Lighting could change!
– People could do weird things!
11
Google Similar Images
http://www.youtube.com/watch?v=6fD2t4d2Ln4
http://similar-images.googlelabs.com/
Systems that learn about the world.
12
Vision: Mathematical Foundations
Differential Geometry
- Probabilistic and Statistical Models

- Fourier Analysis
Extract high-level but low “Eigenfaces” – Princip

dimensional information from low- Component Analysis
level high dimensional data.
Animal Behavior and Census
Bat Tracking:
Collaboration with Biologists
Funded by Office of Naval Research
Demo Video
14
Cell Tracking / Analysis
House et al. – Boston Univer

Linguistic Analysis of Sign Language
on University – American Sign Language Linguistics Research P

Vision Guided Robots
Autonomous
Vehicles
Assistive Robots
Tele-presence Robots Manufacturing

Remote Sensing (Geography)
Gautama et al. – Gent Univer

Computational Neuroscience
Biologically Inspired Vision:
Machine Learning, Artificial Neural Networks
Brain Modelling
Brain-Computer Interfaces
Protein Folding (Biochemistry)
Many Computer Vision techniques used in

computer simulations.
Finance / Machine Learning
Abstract from Bloomberg research talk:

Gary Kazantsev, R&D Machine Learning, 12/05/2013
We will give a brief overview of the machine learning discipline from a

practitioner's perspective and discuss the evolution and development of
several key Bloomberg projects such as sentiment analysis, market impact
prediction, novelty detection, machine translation, social media monitoring
and information extraction . We will show that these interdisciplinary
problems lie at the intersection of linguistics, finance, computer science and
mathematics, requiring methods from signal processing, machine vision
and other fields. Throughout, we will talk about practicalities of delivering
machine learning solutions to problems of finance and highlight issues such
as importance of appropriate problem decomposition, feature engineering
and interpretability.
Human-Computer Interaction
We’re all used to mouse and keyboard…
But you could use a camera to track motion…
Camera Mouse
http://www.cameramouse.org/ (Free Download)
A user with severe

paralysis using the
Camera Mouse
Articles and Videos:

http://www.bu.edu/today/2009/04/10/seeing-eye-mouse
http://www.bu.edu/today/2011/big-meaning-in-the-smallest-
movements/
22
Reading
 http://en.wikipedia.org/wiki/Template_matching
– http://en.wikipedia.org/wiki/Sum_of_absolute_differences
– http://en.wikipedia.org/wiki/Cross-correlation
 http://en.wikipedia.org/wiki/Netpbm_format
 http://en.wikipedia.org/wiki/Pinhole_camera
 http://en.wikipedia.org/wiki/Perspective_projection
 http://en.wikipedia.org/wiki/Camera_matrix
23

Puter Vision

Uploaded by

Copyright:

Available Formats

Puter Vision

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Puter Vision

Uploaded by

Copyright:

Available Formats

CS201 Lecture 02

Recall: Computer graphics in general

Fujifilm camera demo 2

(Some extra detail on the next few slides)

Histogram can be updated

Recall: Video compression uses frame differencing. 6

Sum of Absolute Differences

 Normalized correlation coefficient matching over

– You might not know anything about a scene!

- Probabilistic and Statistical Models

Extract high-level but low “Eigenfaces” – Princip

House et al. – Boston Univer

on University – American Sign Language Linguistics Research P

Tele-presence Robots Manufacturing

Gautama et al. – Gent Univer

Many Computer Vision techniques used in

Abstract from Bloomberg research talk:

We will give a brief overview of the machine learning discipline from a

A user with severe

Articles and Videos:

You might also like