Computer Vision

Instructor: Marc Pollefeys
comp256 Fall 2003

Tuesdays and Thursdays from 3:30-4:45 in SN011

Computer Vision (following Tomaso Poggio, MIT): Computer Vision, formerly an almost esoteric corner of research and regarded as a field of research still in its infancy, has emerged to a key discipline in computer science. Vision companies have emerged and commercial applications become available, ranging from industrial inspection and measurements to security database search, surveillance, multimedia and computer interfaces. Computer Vision is still far from being a solved problem, and most exciting developments, discoveries and applications still lie ahead of us. Understanding the principles of vision has implications far beyond engineering, since visual perception is one of the key modules of human intelligence.

Goal and Objectives
  1. To introduce the fundamental problems of computer vision.   
  2. To introduce the main concepts and techniques used to solve those.
  3. To enable participants to implement solutions for reasonably complex problems.   
  4. To enable the reader to make sense of the literature of computer vision.

Who should attend this course?

Graduate students that are interested to learn about the fundamental concepts of computer vision or desire to use compute vision techniques in their research.  
Prerequisites: Most of the knowledge required should be part of the normal background in Computer Science and undergraduate/graduate Mathematics and Geometry. Comp235 or equivalent is required.

The textbook for this course is "Computer Vision: a modern approach" by David Forsyth and Jean Ponce (should be available at Student Stores).  This is one of the most recent books on computer vision, authored by two very well respected researchers in the field.  This book has become one of the most popular textbooks for computer vision.  

  • Introduction
  • Shape from Stereo, Motion, Shading, ...
  • Segmentation
  • Tracking
  • Recognition

Learning approach
  • Students should preferably read the relevant chapters of the books and/or reading assignements before the class.  
  • In the course the material will then be covered in detail and motivated with real world examples and applications. 
  • There will be assignements to provide students with practical experience of some computer vision techniques. 
  • There will also be a final project where students will solve a real world problem using computer vision techniques.  

Grade distribution
  • Assignments: 40%
  • Class participation: 10%
  • Project  proposal: 10%
  • Final project: 40%
Some usefull links
  • Class 1 introduction
  • Class 2 cameras, lenses and sensors
  • Class 3 Radiometry
  • Class 4 Sources, shading and photometric stereo
  • Class 5 Color
  • Class 6 Linear filters & edges
  • Class 7 Pyramids and textures
  • Class 8 Multiple View Geometry
  • Class 9 Stereo
  • Class 10 Project Proposal Presentation
  • Class 11 Tracking
  • Class 12 Optical flow
  • Class 13 Silhouette and space carving approaches
  • Class 14 Structure from motion (affine)
  • Class 15 Structure from motion (projective)
  • Class 16 Calibration (and self-calibration)
  • Class 17 Segmentation (foreground/background, K-means, graph-based)
  • Class 18 Fitting (Hough, RANSAC, ... )
  • Class 19 Probablistic segmentation (EM and model selection)
  • Class 20 Template matching and object recognition (Classifiers, Neural Nets, SVM,...)
  • Class 21 Recognition by relations
  • Class 22 Range (Structured light, Registration, Merging, Recognition, ...)
Class project

Marc Pollefeys, October 4, 2003.