Abstract:

We describe a unified and transversal framework to teach coding technologies to undergraduate students of computer science as well as multimedia and communication engineering. We focus on a set of three courses in the Degree on Audiovisual System recently offered by a Spanish university, which follows the Bologna declaration for European Higher Education.

Coding the information with the minimum bit rate possible is a problem that was addressed by Shannon in his Source Coding Theorem [Cover & Thomas, 1991]. The limit is given by the entropy of the information source. However, different multimedia sources suffer from the presence of different kinds of redundancy, and approaching to the entropy limit becomes a different challenge depending on the specific data source. In image coding, the redundancy can be spatial, perceptual or statistical. In video coding, there is also temporal redundancy. In audio coding, the redundancy can be either temporal, perceptual or statistical. We propose a set of contents that are both theoretical and practical, designed to provide a coherent point of view to the field. The students can relate the contents acquired in a given course to the ones studied in previous or later courses.

A unit on Information Theory is taught in a course on Digital Communication Theory (2nd undergraduate year). There, channel coding is paid more attention than source coding, because the course is more concerned with transmission than with coding. However, two sessions, one of them taking place in the lab, are devoted to introduce the students to the basics of source coding for reducing the statistical redundancy of an information source. In the session devoted to theory they study Huffman and run-length codes. They connect this principles with Zipf and Benford laws. In the practical session, they are given the task of compressing text by exploiting the statistics of natural language.

In the Digital Image Processing course (3rd year), there is a coding unit in which the students learn the different sources of redundancy (spatial, temporal, perceptual and statistical) in digital image and video. In the practical sessions, they study a simplified version of a standard coder (JPEG) for image coding, and perform basic movement estimation for video coding [Clarke, 1995].

In the Digital Audio Processing course (4th year) the students study in one of the units the basics of temporal redundancy reduction by linear prediction, and specific source coding algorithms for super-gaussian distributions, which is the case in audio sources. Also, they are introduced the principles of perceptual audio coding. In the practical sessions, the students implement an Audiopak coder, which consists of a lossless coder with a code-based linear prediction and a Golomb-Rice source coder [Hans & Schafer, 2001].

References

T. M. Cover and J. A. Thomas, “Elements of Information Theory”. New York: Wiley, 1991.

J. Clarke: “Digital Compression of Still Images and Video”. London, UK. Academic Press;1995.

M. Hans and R. Schafer, “Lossless Compression of Digital Audio”, IEEE Signal Processing Magazine, pp 21-32, July 2001.

Keywords:

Bologna Declaration, Coding Technologies, Cross-curricular contents.