Syllabus CS676 Jan-Apr 2011

CS676 Computer Vision and Image Processing

Instructors: Amitabha Mukerjee (IITK) and Prithwijit Guha (TCS labs Delhi)

The first seven weeks of the course, will be intensive with one homework each week, mostly based on openCV. This will be followed by a major project over the next six weeks. Projects will be single-student efforts.

Course timings: Mon Tue Fri 11:00-12:00 CS101.

Syllabus

Each topic will be covered in roughly one week.

1. Image Formation and Coordinate Transformations
   Camera Matrix, Motion/Stereo Pin-hole model,
   Human eye / cognitive aspects of colour / 3D space; illumination;
   Sampling and Quantization
   Coordinate transformations and camera parameters

   HW Jan3-10: 3D transformations problem (theory); 
   Introduction to OpenCV : Image Data Structure, Coding format,
      

   Reading: Chapters 1 and 2 and parts of ch 3 of FP. 

2. Image Processing - Noise Removal, Blurring, 
        Edge Detection: Canny / Gaussian/ Gabor/Texture Edges/ 
	Curvature / Corner Detection.  
   
   

   Motion Estimation : Horn-Schunk Optical Flow Formulation
       Euler-Lagrange formulation : Calculus of variations theory. 
       Structure Recovery from Motion [Kanade]
   HW: flow homework 

   Reading: (tentative) Chapters 7-8 and parts of ch 9 of FP. 

3. Segmentation - Concept of Figure vs. Ground, Watershed, 
   Change Detection, 
   Background Subtraction, Texture Segmentation
   Gaussian Mixture Models - Applications in Color/Motion based Image
   Segmentation, Background Modeling and Shape Clustering

   HW: Change Detection; Bacgkround subtraction

   Reading: TBA

4. Machine Learning techniques in Vision
   Bayesian Classification, Maximum Likelihood Methods, 
   Neural Networks; Non-parametric models; Manifold estimation
   Support Vector Machines ; Temporal sequence learning
   HW: Classifiers for edges; object categories

   Reading: Chapters 1 and 2 of Bishop; additional material TBA

5. Introduction to  Object Tracking - Exhaustive vs. Stochastic Search 
   Shapes, Contours, and Appearance Models.  
   Mean-shift tracking; Contour-based models

6. Object Modeling and Recognition
   Fundamental matrix / Epipolar geometry
   Adaboost approaches: Face Detection / Recognition
   Large Datasets; Attention models.  
   HW: Haar-based modeling of object classes

7. Applications: Surveillance, Object detection, etc. 

In addition, a number of advanced topics such as activity modeling and recognition, cognitive aspects of vision, robot self-localization, etc. will be covered by case presentations during the latter part of the course.

Projects

Project topics would need to be decided within approximately five weeks into the course, and there would be several intermediate checkpoints.

All projects will be done alone.

Possible topics:

 camera calibration
 CRF segentation
 object modeling
 activity recognition
 image manifolds
 PHOW 
 Le Cun models  

Grading Policy: (tentative)

Exams: 25-30% Assignments 30-40% Projects 40-50%

Texts

A. [FP]: David Forsyth and Jean Ponce, Computer Vision: A modern Approach, Prentice
   Hall India 2004: 
B. Christopher Bishop, Pattern Recognition and Machine Learning, Springer, 2008

OTHER REFERENCE TEXTS:

1. E.R. Davies, Machine Vision, Theory Algorithms Practicalities, Elsevier
   2005
2. Milan Sonka, Vaclav Hlavac and Roger Boyle, Image Processing, Analysis,
   and Machine Vision.  Brooks/Cole / Thomson 1999
	Basics of some image processing aspects. Texture
3. Chapter 24 (Perception) of Russell and Norvig: AI: A modern Approach,
   Prentice Hall 2000.
4. Richard Hartley and Andrew Zisserman, Multiple View Geometry in Computer
   Vision, Cambridge Univ Press 2000
	More detailed treatment of 3D structure recovery
5. Richard O. Duda, Peter E. Hart, and David G. Stork, Pattern
      Classification, 2nd ed., Wiley Asia, 2002