The Wavelet Digest Homepage
Return to the homepage
Search the complete Wavelet Digest database
Help about the Wavelet Digest mailing list
About the Wavelet Digest
The Digest The Community
 Latest Issue  Back Issues  Events  Gallery
The Wavelet Digest
   -> Volume 1, Issue 4

Ph.D. Thesis alert
images/spacer.gifimages/spacer.gif Reply into Digest
Previous :: Next  
Author Message
Touradj EBRAHIMI, Swiss Federal Institute of Technology at Lausanne.

PostPosted: Sat Aug 22, 1992 10:41 pm    
Subject: Ph.D. Thesis alert
Reply with quote

Ph.D. Thesis alert

Title: Perceptually derived localized linear operators
- Application to image sequence compression -

by : Touradj EBRAHIMI

Directed by: Professor Murat Kunt
Signal Processing Laboratory
Swiss Federal Institute of Technology at Lausanne (EPFL)
CH-1015 Lausanne

date: June 12th, 1992


The theory of linear operators covers a large number of applications
in applied mathematics and signal processing. This thesis discusses
three commonly used linear operators, namely, the short time Fourier
transform, the wavelet decomposition and the subband decomposition.
Common features and differences between the above operators are
examined in the frame work of image sequence compression. In
particular, the Gabor expansion, as a special case of short time
Fourier transform in discussed in details. A new and efficient
technique to perform Gabor expansion of signals is introduced. A
comparative study between the above approach and other commonly used
techniques for Gabor expansion of signals is given in terms of
computational complexity. The subband decomposition of signals and the
wavelets theory are two sub-classes of linear operators which were
developed in two different disciplines. However, there exist big
similarities between the two approaches. After a brief introduction of
subband decomposition and wavelet theory, the equivalence and
differences between them are discussed, in both orthonormal and
biorthonormal cases. The multiresolution representation of data is
considered in particular. A new approach to filter bank design in
time domain, using matrix formulation is introduced. Using this
approach, an efficient filter bank with low complexity and good
frequency response is designed. The characteristics of this filter
bank are compared with that of other commonly used short kernel filter
banks, for video compression applications. The generic coding is an
important component in any modern communication system. In this
thesis, a generic video codec is introduced which is able to compress
efficiently image sequences with any format at desired bitrate,
without any change in its architecture. The approach is based on
multiresolution representation of data, which is generated by the
filter bank proposed in this work. The use of multiresolution data
structure in conjunction with other components of the system such as
multiresolution motion estimation, allows a simple and efficient
implementation. Simulations on typical image sequences show that it is
possible to perform generic coding with reduced complexity and good


Copies of this thesis can be obtained by requesting from the
above address. A postscript version of the same document (with lower
quality of pictures) can also be obtained from ``".

Swiss Federal Institute of Technology at Lausanne
Signal Processing Laboratory
1015 Lausanne L TTTTTTT SSSSSS
Switzerland L T S
tel: (+4121) 693 2796 L T S
fax: (+4121) 693 4660 LLLLLL T SSSSSS
All times are GMT + 1 Hour
Page 1 of 1

Jump to: 

disclaimer -
Powered by phpBB

This page was created in 0.027641 seconds : 18 queries executed : GZIP compression disabled