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 2, Issue 14

images/spacer.gifimages/spacer.gif Reply into Digest
Previous :: Next  
Author Message
Vinod Raghavan, School of Chemical Engineering, Oklahoma State University

PostPosted: Fri Nov 29, 2002 3:36 pm    
Subject: Questions
Reply with quote


Dear Waveletters,

I need lots of help. First I want to know how to pad a finite length
signal before the decomposition. I have tried the circular convolution
technique and for the Daubechies family of wavelets, it works fine. The
problem creeps up when I am dealing with a signal like a step function.
The decomposition coefficients come out real wierd, more so when I go
down a number of levels. But, if I repeat the same padding procedure
while reconstructing it, I get perfect reconstructions.
My problems is that I am using the decomposition coefficients for Pattern
Recognition purposes and can't afford to get poor coefficients. Perfect
reconstruction is secondary to me , but is important. I tried padding it
using Mallat's procedure in " A theory for multi-resolution signal
decomposition: a wavelet representation" in IEEE Transactions on Pattern
Recognition and Machine Intelligence,Vol.11, No.7 July 1989, but then
the decomposition somehow screws up. Does any one else have this same
What kind of padding should I use to avoid poor coefficients and
still get perfect reconstruction?
I think this is a critical issue, but none of the articles I have read talk
about this. Have I missed any article?

My second problem: Is there any article in English that deals with the
construction of the Meyer,Battle-Lemarie and other wavelets(other than the
Daubechies and the Biorthogonal wavelets developed by Daubechies) explicitly?
Is there any article where the coefficients for these wavelets are tabulated
for some orders so that I can compare my values with these. As far as I
know only Daubechies has done that. That benefits non-mathematicians like me
a lot.

I would appreciate any help.

Vinod Raghavan.
423 Engineering North, School of Chemical Engineering
Oklahoma State University, Stillwater,OK 74078 USA
All times are GMT + 1 Hour
Page 1 of 1

Jump to: 

disclaimer -
Powered by phpBB

This page was created in 0.518203 seconds : 18 queries executed : GZIP compression disabled