Course Notes

The lecture notes are distributed in printed form; if you miss them being handed out in the first lecture, you may be able to obtain one of the spares from CSEE's General Office. You can also read through the notes online; this includes corrections for any errors we might spot in them during the term.

Each chapter of the notes has an associated quiz, which is normally done at the start of the lecture after which that material has been covered; they are made available here after they have been done in lectures. You can download these quizzes from the list below and work through them on your own computer. If you use Adobe's Acrobat Reader to view the questions, you'll be able to have it mark them and give you your score.

0. How much do you know before the course starts?
1. Introduction
2. Simple image processing operations
3. Convolution
4. Feature Detection
5. Feature Matching and the Hough Transform
6. Tracking
7. Depth Recovery from Stereo
8. Experimentation and Evaluation

Assignments

There are four short assignments associated with EE322. These are pencil-and-paper (-and-calculator) exercises, done during the four problem classes associated with the module. They ensure that you have learnt the essential skills of computer vision without making you spend hours and hours doing (say) programming exercises. Every person receives a different 'image' which they will work on, and I normally mark and return the assignments within a couple of days to ensure you receive prompt feedback.

Progress test

There is a progress test associated with this module. It will take place in week 8, along with all the other progress tests.

Trying things out in practice

There is a wealth of computer vision software Out There, much of it intended for research use. Some general rules of thumb are:

• If you do a project that involves programming computer vision techniques or applications, your best bet is to investigate the OpenCV library. This is intended for online, real-time use, so the various functions and operators run quickly but the source code is often difficult to understand.
• If you want to get a feel for what image processing operators do, please have a play with ImageJ, which should be installed on the machines in our software laboratories. ImageJ is written in Java and hence runs on Windows, Linux and MacOS; it is GUI-driven and has a fair selection of image processing features that you can try out. It is actually intended for biomedical researchers and has a number of powerful features that allow its capabilities to be extended and enhanced, though we shan't explore them in EE322.
• To give you an idea of how some computer vision operators may be programmed, I shall shortly put on this website EVE, an implementation of some fundamental image processing operators in (numerical) Python. I actually use EVE on real problems as it's useful for prototyping; but I suggest you treat it as a kind of pseudo-code to see how operators are implemented.