Applied Analysis : Fourier analysis, data compression (MPEG, JPEG), wavelets

MAT 4996/5326

January 2000


Description

Audio and video signal processing is the backbone of modern digital communication. It all depends on fast transforms, be it the fast Fourier, the fast cosine, or the fast wavelet transform. The windowed Fourier transform and the windowed cosine transform localize the signal in time and frequency. The wavelet transform localizes the signal in time and scale. It is the discrete versions of these transforms that are used in practice, and each of them admits a fast implementation. No single transform has all the good points and none of the bad points. The "Joint Photographic Experts Group" (JPEG), the "Joint Bi-level Imaging Group" (JBIG) and the "Moving Picture Experts Group" (MPEG) of the International Standard Organization (ISO) have adopted the windowed cosine transform as the basic tool to localize, quantize and compress the signal. JBIG allows for a greater compression ratio than JPEG. JPEG part 1-3 has a set of parameters now called JEPG utilities. The recently agreed standard JPEG part 4, called JURA for "JPEG Utilities Registration Authority", provides for extending the standard by registering new utilities defined for the use of individual JPEG users. JPEG 2000 is proposing wavelets to greatly reduce ringing artifacts present with JPEG and JBIG. A pdf document on JPEG 2000 entitled JPEG-2000 : Background, scope, and technical description by Majid Rabbani, can be downloaded . A widely used audio coding format known as MP3 (aka MPEG Audio Layer 3) will be considered. One immediately sees the impact of the decisions of the expert groups on hardware and software design for high definition television, medical applications, and so on. Among the many applications of wavelets stands finger print compression by the FBI.

The first lecture will skim a few aspects of MPEG and JPEG, taken from ref. no. 3 below: Digital video : an introduction to MPEG-2 . A few lectures will be devoted to continuous and discrete Fourier and cosine transforms. It is expected that the course, in the main, will keep close to the text of Strang and Nguyen.

Instructor: Rémi Vaillancourt,

Prerequisites: Third year engineering mathematics or third year analysis for mathematics and science.

Course material: The book, in stock at the Unversity Bookstore, Wavelets and filter banks of Strang and Nguyen, referenced below, will serve as the basic textbook for the course. Theoretical and practical assignments and projects will be part of the course.

Evaluation and grading scheme: Assignments, projects and an open book final examination. Following engineering practice, at least 50 percents of the final grade will be for work in a controlled environment.

Classroom and schedule: Lectures will be held in KED B05 at 585 King Edward Ave. on Mondays and Wednesdays from 19:00 till 20:30. First meeting on 5 January 2000. Rescheduling may be considered if desirable and possible (the rule is unaninous approval of any change).

Tentative office hours: Tuesdays and Thursdays from 16:00 to 17:00 and on appointment. e-mail is the preferred medium. Telephone is also possible.

References:

  1. A comprehensive text for JPEG is JPEG still image data compression standard by William B. Pennebaker and Joan L. Mitchell, Van Nostrand Reinhold, New York, 1993.
  2. A text which surveys standards in chapters 7 to 10 is Digital pictures : representation, compression, and standards by Arun N. Netravali and Barry G. Haskell, 2nd. ed., Plenum Press, New York, 1995.
  3. A text which surveys MPEG-2 is Digital video : an introduction to MPEG-2 by Barry G. Haskell, Atul Puri, and Arun N. Netravali, Chapman & Hall, London, 1997.
  4. A text which explains the basics behind the standards without bearing on the details is Image and video compression standards : algorithms and architectures by Vasudev Bahskarna and Konstatinos Konstantinides, Kluwer Academic Publishers, Boston, 1995.
  5. A text for wavelets that uses both the signal processing and mathematics languages is Wavelets and filter banks by Gilbert Strang and Truong Nguyen, Wellesley-Cambridge Press, Box 812060, Wellesley MA 02181, 1996.
  6. A reference on wavelets from an enginnering point of view with a solid mathematical foundation is Wavelets and subband coding by Martin Vetterli and Jelena Kovacevic, Prentice-Hall, 1995.
  7. A graduate course on wavelet is A wavelet tour of signal processing by Stéphane Mallat, Academic Press, San Diego, 1998.

Relevent web pages:

Last modified: 24 December 1999.