COMP30432: Computer Vision (2009-2010)
To provide a broad introduction to Computer Vision and Image Interpretation. To introduce the basic concepts and algorithmic tools of Computer Vision To explore the importance of modelling and representation in interpretation of images. To provide an understanding of the range of processing components involved in image interpretation systems.
A student completing this course unit should:
Have an understanding of the theoretical and practical capabilities of Computer Vision. (A)
Have a knowledge of common Computer Vision and Image Interpretation algorithms. (A)
Have an understanding of the design of algorithms. (A and B)
Be able to formulate solutions to problems in Computer Vision. (A and B)
Assessment of Learning outcomesLearning outcomes (1), (2), (3) and (4) are assessed by examination.
Contribution to Programme Learning OutcomesThis course contributes to learning outcomes; A1, A2, A5, B1.
The tools and algorithms of computer vision are introduced in the context of two major capabilities required of visual systems: recognition - finding and identifying expected things in images and 3D interpretation - understanding a dynamic 3D scene from 2D images or sequences of images. These capabilities are explored using applications of varying levels of complexity: recognising man-made objects, interpreting medical images, face recognition, robotics, scene reconstruction and surveillance.
Students taking the module will require some basic familiarity with matrix and vector algebra, such as that covered in MT1662 or MT1672. Tutorial material will be provided where the mathematics goes beyond the scope of those modules. A general familiarity with basic concepts of calculus (integration, partial differentiation) will also be useful.
Introduction:The role of Computer Vision, applications, successes, research issues; its relationship to natural vision, basic image properties.
Image Interpretation: Finding Things in Images
Exemplars: Face Recognition, Medical Image Analysis, Recognising man-made objects.
Representing knowledge of expected image contents.
Matching models to image data.
Simple implicit models: Thresholding, Edge finding, Organising mechanisms. Mathematical Morphology.
Hough Transform, Template Matching (correlation), Active Contour Models (Snakes).
Multi-resolution approaches, Feature-based models, Edges, Corners, Object Recognition.
Flexible Templates, Search, Constraining Model Fit, Statistical Models, Describing natural variability, Active Shape Models, Recognition, Classification.
Beyond 2D: 3D and Motion
Exemplars: Robot Vision, Scene Reconstruction, Surveillance.
Stereopsis: Recovering depth, The correspondence problem, Stereo constraints.
Motion: Tracking, Image differencing, Feature matching Models, Optic flow.
In addition to the material in lecture notes and textbooks, Self-test questions and solutions will be provided. For some topics, practical exercises, with associated MATALB scripts and images will be available for use by students unsupervised. These additional materials may be downloaded from the course web site.
Jain and Sonka will provide most of the reading material for the course. The scope of both goes well beyond the course content. Selected passages will form essential background reading, supplemented by course notes. Students should have access to copies of both.
Low may provide useful background on some elementary material for the module.
Gonzalez and Woods is a standard image processing text. It is very good on the mathematical basis of pixel-pushing, but does not cover interpretation.
Where necessary, the recommended material in Sonka or Jain will be supplemented by handouts.
Core TextTitle: Machine vision
Author: Jain, Ramesh and Rangachar Kasturi and Brian G. Schunck
This is the main text, covering most, but not all material in the course
Supplementary TextTitle: Image processing, analysis and machine vision (4th edition)
Author: Sonka, Milan and Vaclav Hlavac and Roger Boyle
Publisher: Nelson Engineering
Covers more material than Jain et al. Much more mathematical in approach.
Supplementary TextTitle: Introductory computer vision and image processing
Author: Low, Adrian
An easy read introducing some of the early course material and some other elementary topics
Supplementary TextTitle: Digital image processing (3rd edition)
Author: Gonzalez, Rafael C. and Richard E. Woods
Classic text in image processing. Not much on interpretation.