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 7, Issue 5


Thesis: Optimal Perfect Reconstruction Filter Banks
 
images/spacer.gifimages/spacer.gif Reply into Digest
Previous :: Next  
Author Message
ilangko@tele.ntnu.no (Ilangko Balasingham )
Guest





PostPosted: Wed May 06, 1998 3:55 pm    
Subject: Thesis: Optimal Perfect Reconstruction Filter Banks
Reply with quote

#2 Thesis: Optimal Perfect Reconstruction Filter Banks

Title: On Optimal Perfect Reconstruction Filter Banks
for Image Compression

Author: Ilangko Balasingham <ilangko@tele.ntnu.no>
Address: Department of Telecommunications
Norwegian University of Science and Technology
N-7034 Trondheim, Norway
Advisor: Professor Tor A. Ramstad <tor@tele.ntnu.no>

Compression of images have important applications in storage,
transmission, and processing. One of the modules in a compression
scheme is the filter bank. This thesis focuses on selection,
optimization, and frequency partitioning of filter banks for both
lossless and lossy image compression. All filter banks constructed in
this thesis belong to the biorthogonal, perfect reconstruction class.

The wavelet transforms have received much attention lately. The
connection between wavelet transform and filter bank is
investigated. It is known that the discrete wavelet transforms is a
sub-class of filter banks. In the thesis it is shown that by relaxing
the maximum regularity constraint, usually imposed on wavelet
transforms, and instead of introducing a smoothness constraint for
synthesis lowpass filter, higher coding gain in terms of PSNR and
visual quality is obtained. This is demonstrated for a (9/7) system
constructed on such premises which outcompetes the popular 9/7 wavelet
transform. Hence, the maximum regularity constraint is irrelevant.
The proposed subband coding gain optimized filter banks are among the
best in the literature.

Frequency domain tiling is scrutinized, and it is found that very good
coding results are obtained by using an 8 by 8 parallel filter bank
where the lowpass-lowpass band is further split by a dyadic filter
bank. The obtained coding results are among the best published so far
the literature.

A novel way of constructing integer coefficient filter banks is
proposed for lossless image coding. Discrete search of integer filter
coefficients under constrained optimization is used as a tool. The
compression ratios for a number of images indicate that the tested
filter banks behave differently due to local variation of the image
statistics, and therefore a clear winner is difficult to choose.
However, a dyadic split using 2-channel 17/11 and 5/3 filter banks
have satisfactory performances. The compression ratios are comparable
to state-of-the-art lossless schemes when incorporating arithmetic
coder allocation as the compression engine.

A copy of the thesis can be obtained from
http://www.tele.ntnu.no/~ilangko/thesis.html.
All times are GMT + 1 Hour
Page 1 of 1

 
Jump to: 
 


disclaimer - webmaster@wavelet.org
Powered by phpBB

This page was created in 0.026740 seconds : 18 queries executed : GZIP compression disabled