Background: Currently we have 2 courses on the books:
	COMP 9517 Image Processing and Applications (contains topics from
		IP, CV and PR, but a bit dated)
	COMP 9518 Pattern Recognition and Computer Vision (on PR and CV,
		but IP needs to be introduced as basic technology)
EE also has a course on Image Processing.

Aim: A new course that will combine topics from 9517 and 9518 into a
single course. To remove the need to take 2 courses, and modernise
contents. The proposal contains elements of image processing, computer
vision and pattern recognition.

Pre-requisite: any 2 level 3 COMP/SENG courses
Suggested Topics

This is based on the recommendations in the most modern text available:
David Forsyth and Jean Ponce, Computer Vision a modern approach, Prentice
Hall, 2003. I have also cross-checked with current contents of 9517 and
9518.
(The authors of the text are affliated with University of California,
Berkeley and University of Illinois, Urbana-Champaign, respectively)

1. Introduction: pinhole camera, radiometry
2. Sources, shadows, shading: local shading models, point, line and area
        sources, photometric stereo
3. Colour Vision: physics of colour, colour perception, representation,
	model for image colour, surface colour from image colour
4. Linear filters: smoothing, edge detection using convolution, fourier
	transform, sampling and aliasing, normalised correlation, scale
	and image pyramids
5. Texture: statistics of filter outputs, synthesis, shape from texture
6. Geometry of two views, stereopsis
7. Segmentation by clustering: grouping, clustering pixels,
	graph-theoretic clustering
8. Segmentation by model fitting: Hough transform, fitting lines, curves,
	robustness
9. Segmentation and filtering by probabilistic methods: EM algorithm
10. Tracking with linear dynamic models: Kalman filters, data association
11. Camera calibration: geometric camera models, calibration
12. Model-based Vision: hypotheses formation, verification
13. Template matching using classifiers: building classifiers, feature
	selection, neural networks, Support Vector Machines
14. Recognition by relations between templates: relational reasoning. HMMs
15. Applications