This module aims to provide a comprehensive and rigorous treatment of the main approaches in multimedia (document, image, video, graphics) analysis. Three main themes are covered:
representation and modeling of multimedia entities using various modeling approaches,
matching and comparison of a model with an input entity, or between two entities, and
derivation of a model from sample entities.
It focuses on the non-vector-space approach, which complements the vector-space approach to multimedia analysis.

In addition to covering the main approaches in multimedia analysis, this module also aims to teach a rigorous approach to problem solving through the theoretical treatment and practical illustration of the methods. This balanced approach provides both a solid foundation and the necessary practical tools for the students to effectively engage in multimedia research. Moreover, the methods that they learn can extend well beyond the domain of multimedia into other computer science areas. Such application examples will also be highlighted in this module.
 

After taking the module, a student should be able to perform the following research tasks:
Write a formal statement of the definition of a multimedia analysis problem.
Apply one or more of the approaches to devise a method to solve the problem.
Develop performance indices that can be computed to measure the performance of the
    method devised to solve the problem.
Evaluate and analyze the performance of the method.

We will focus mainly on images, 3D models, and videos.
 

Mathematics is an indispensable tool for multimedia analysis. Unfortunately, it has presented itself as a major obstacle for beginners in this exciting area. Our philosophy in teaching this module is

Our aim is simple: Help the students understand and master the tools of mathematics instead of confusing everybody with endless proofs.
 

syllabus
 

Knowledge in CS4243 Computer Vision, CS3241 Computer Graphics, CS5341 Computational Photography or CS5240 Theory and Practice of Multimedia will be very helpful for this course.

Note that if you are a graduate student, you can still register for this module if you don't have the prerequisite. But, you are expected to be familiar with the materials covered in the prerequisite. If you are not familiar with them, then you should learn the materials yourself before taking this module.

You should be familiar with using either Matlab or C/C++ libraries for problem solving. Check resource webpage for a list of useful software and libraries.

Check here for sample numerical algorithms that are often used to solve multimedia analysis problems.
 

schedule, topics, lecture materials
 
Items
Weightage
2 problem definition assignments
25%
1 problem solving assignment
15%
1 paper presentation 20%
1 individual project
40%
Total
100%

Here are some exciting projects that the students have completed in the past.

Assignment grades
 
It would be nice to have a textbook. Unfortunately, nobody has written one on such wide ranging topics. Conference and journal papers will constitute most of the reading materials, and will be listed in the lecture notes.
 

Software, algorithms, libraries
Examples and video demonstrations of algorithms
Very brief introduction to various numerical algorithms that are frequently used to solve
    multimedia analysis problems.
  

A/Prof. Leow Wee Kheng

Last updated: 7 Jan 2009