CS636 Computer Vision

Revised by Christopher Jaynes

Credits: 3


Course Description:

This course covers both the fundamental issues and advanced techniques for
the acquisition and processing of visual information.  The course will
focus on the automatic interpretation of computer images to support 
a variety of intelligent tasks such as navigation, image retrieval, 
recognition of objects and object behavior, and scene understanding.  Students
will be introduced to techniques that support these tasks including camera
calibration and photogrammetry, stereo correlation, surface representation, 
active contours, geometric hashing, computation of optic flow, model 
fitting, and classification methods.  Related issues in knowledge 
representation, search, learning, and general scene understanding systems 
will be covered.

Prerequisites: Either CS5-- Situated Computing, or CS635, or consent of
 instructor.

Required Skills:

Students should have a solid capability in linear algebra, and a familiarity
with numerical techniques and geometry.  An understanding of the issues in 
computer vision (acquired through the "Situated Computing" course) or 
a background in digital image processing (acquired through the CS635 course)
is encouraged.

Learning Outcomes:

Students will learn both the fundamental issues and advanced techniques 
related to computer vision and knowledge-directed processing of digital images.
The primary concepts learned will be camera calibration and photogrammetry, 
stereo correlation, surface representation, active contours, geometric hashing, 
computation of optic flow, and model fitting.  In addition, students will 
understand the key concepts in the engineering of a knowledge-directed vision
system, such as representation of belief and uncertainty, schemas, and truth
maintenance systems for computer vision.  Finally, students will understand 
how these techniques and concepts fit together in successful computer
vision systems that support "intelligent" tasks, such as path planning, 
navigation and road following, object recognition, learning, and scene 
classification.

	

Course Content:

	* Introduction to Computer Vision
		-Image formation, Image geometry
		-Image Processing (feature enhancement/detection)
		-Ill-posed vision problems, knowledge-based constraints
		-Color Perception

	* Computer Vision Topics
		-Pose determination, registration
		-Motion Analysis: Structure from motion
		-Photogrammetry, Calibration
		-Surface Modeling: Generalized Cylinders, Superquadrics
		-Range Segmentation
		-Appearance Based Indexing
		-Scale Space, Steerable Filters
		-Regression Analysis (Surface Fitting, refinement)
		-Geometric Hashing
		-Scene/Object Classification
		
	 * Knowledge-Based Computer Vision
		-Knowledge Representation
		-Uncertainty and Belief
			Bayesian Networks
		-Perceptual Organization
		-Spatio-Temporal Grouping
		-Intelligent Control
		-Motion Understanding, Behavior Modeling (Stochastic)
		-Decision Modeling, Utility Theory
		-Learning and Vision
		       
	  * Active Vision
		-Gaze Control, Visual Search
		-Cooperative Vision (multi-agent)

Textbooks:

 "High-Level Vision: Object Recognition and Visual Cognition" Shimon
 Ullman.  MIT Press, 1992.  ISBN: 0262210134.

 "