Tutorial TS3
Data Compression for Multimedia

Half day, Level Intermediate

Lecturer

(Note: this tutorial was initially planned with Prof. Amer as a lecturer. Unfortunately, for personal reasons, Prof. Amer had to cancel his participation)

Professor Jean-Luc Dugelay
Multimedia Communications Department
Institut EURECOM
BP 193
06904 Sophia-Antipolis cedex, France
Tel: +33 (0)4 93 00 26 41 office
Fax: +33 (0)4 93 00 26 41
Email: dugelay@eurecom.fr
Web site: http://www.eurecom.fr/~image/

Abstract

This tutorial is targeted at the practically oriented researcher or developer who is interested in details about the foundation concepts and mechanisms used in multimedia data compression.

An exponentially growing amount of data, image, video, music, voice, virtual reality, etc., is being transmitted around the world. No matter how much we increase telecommunication bandwidth and disk storage capacity, a vital practical need remains to compress data so that multimedia can be transmitted faster and stored more efficiently. Data compression reduces the size of multimedia by reducing an object's redundancy.

Lossless techniques preserve an original file bit for bit after compression and decompression. Lossy techniques obtain significantly greater compression than lossless, but at a penalty of changing or distorting the original file. Lossy compression investigates the tradeoff of compression vs. error distortion.

At the University of Delaware, Professor Amer teaches a graduate level computer science course on data compression. This half-day tutorial summarizes the highlights of the course thereby giving each student a thorough introduction to today’s important approaches to data and multimedia compression. This tutorial culminates with an explanation of JPEG2000 and MPEG, both of which encompass multiple compression methods: lossless and lossy.

Detailed outline

Introduction
- History of data compression
- Methods vs. software vs. standards
- General Characteristics
  - Symmetric vs. asymmetric
  - Uniform vs. nonuniform
  - Types of redundancy
  - Lossless vs. lossy
Lossless compression
- Metrics - how to evaluate lossless compression
- Statistical coding - Huffman
- Dictionary coding - Lempel-Ziv-Welch
- Run-length coding - facsimile
- Arithmetic coding
Lossy compression
- Metrics (SNR, PSNR) - how to evaluate lossy compression
- Scalar quantization
- Vector quantization
- Differential methods (e.g., for speech/audio)
  - Examples: DPCM, ADPCM
- Subband/transform methods (e.g., for images/video)
  - Example: discrete cosine transform
  - Example: wavelet
Image/Video Standards
- GIF, JPEG, JPEG2000
- MPEG1, MPEG2, MPEG4, MPEG7

Lecturer's Biography

Jean-luc DUGELAY was born in Rouen, France, in 1965.

From 1989 to 1992, he worked for France Telecom Research (formerly CNET - CCETT) He received the Ph.D. degree in Computer Science in 1992 from Rennes University.

He is currently with the Institut EURECOM, Multimedia Communications dept., Sophia Antipolis, France, as a Professor; and with the University of California, Santa Barbara, ECE dept., SCL Lab., as a Visiting Researcher.

His research interests currently include Image Processing and Coding, Watermarking and Indexing, Video Communications, Virtual Reality and 3D Imaging.