[Home]
[Back Issues]
[Search]
[Vol 1]
[Vol 2]
[Vol 3]
[Vol 4]
[Vol 5]
The Wavelet Digest - Volume 2, Issue 16 (November 10, 1993)
Today's Editor: Wim Sweldens
sweldens@math.sc.edu
Today's Topics:
1. Short course: Wavelet Electrodynamics
2. Preprint available: Application of wavelets to EEG
3. CFP: Visual Communications and Image Processing '94 (VCIP '94)
4. Looking for code in Daubechies book
5. Preprint available: Portrait of frames
6. Question: Wavelets for Image Compression
7. Computing line integrals of a 3D function described in wavelets?
8. Papers available:
9. Short course on wavelets filter banks and applications
Submissions for Wavelet Digest:
E-mail to wavelet@math.sc.edu with "submit" as subject.
Subscriptions for Wavelet Digest:
E-mail to wavelet@math.sc.edu with "subscribe" as subject.
To unsubscribe, e-mail with "unsubscribe" followed by your e-mail
address as subject. To change address, unsubscribe and resubscribe.
Archive site, preprints, references and back issues:
Anonymous ftp to maxwell.math.sc.edu (129.252.12.3),
directories /pub/wavelet and /pub/imi_93.
Gopher and Xmosaic server: bigcheese.math.sc.edu.
Current number of subscribers: 3004
From: jodi@image.mbvlab.wpafb.af.mil (Jodi Nix)
Subject: Short course: Wavelet Electrodynamics
March 21, 1994 there will be a one day short course on Wavelet Electrodynamics
offered in conjunction with the Applied Computational Electromagnetics Society
(ACES) 1994 Symposium. The course and 10th Annual symposium will be at the
Doubletree Hotel and Convention Center in Monterey, CA. Course description:
Maxwell's equations are symmetric under space-time translations and dilations.
This is used to write an arbitary solution as a superposition of
electromagnetic wavelets. Possible applications include:
(a) Designing electromagnetic waves with initial data specified locally
and by scale.
(b) An economic description of electromagnectic pulses emanating from moving
or accelerating objects.
The course will consist of four 90 minute lectures. Notes will be provided.
Reference: G. Kaiser, A Friendly Guide to Wavelets, Birkhauser,
available 1/94.
If you would like additional information on the course and/or symposium please
contact Jodi Nix at 513-476-3550 or email: jodi@image.mbvlab.wpafb.af.mil
From: trejo@nprdc.navy.mil (Leonard J. Trejo)
Subject: Preprint available: Application of wavelets to EEG
Preprint available -- Application of wavelets to feature extraction of EEG
(Request pre-print by e-mail message to trejo@nprdc.navy.mil)
Trejo, L. J. and Shensa, M. J. (1993b, in press). Linear and neural
network models for predicting human signal detection performance from
event-related potentials: A comparison of the wavelet transform with
other feature extraction methods. Proceedings of the 1993 Simulation
Technology Multiconference, San Francisco, CA, November 7-10. San
Diego: Society for Computer Simulation.
ABSTRACT
This report describes the development and evaluation of mathematical
models for predicting human performance from discrete wavelet
transforms (DWT) of event-related potentials (ERP) elicited by
task-relevant stimuli. The DWT was compared to principal components
analysis (PCA) for representation of ERPs in linear regression
and neural network models developed to predict a composite measure
of human signal detection performance. Linear regression models
based on coefficients of the decimated DWT predicted signal
detection performance with half as many free parameters as
comparable models based on PCA scores and were relatively more
resistant to model degradation due to over-fitting.
Feed-forward neural networks were trained using the backpropagation
algorithm to predict signal detection performance based on raw ERPs,
PCA scores, or high-power coefficients of the the DWT. Neural
networks based on high-power DWT coefficients trained with fewer
iterations, generalized to new data better, and were more resistant
to overfitting than networks based on raw ERPs. Networks based on
PCA scores did not generalize to new data as well as either the DWT
network or the raw ERP network.
The results show that wavelet expansions represent the ERP
efficiently and extract behaviorally important features for use in
linear regression or neural network models of human performance.
The efficiency of the DWT is discussed in terms of its decorrelation
and energy compaction properties. In addition, the DWT models
provided evidence that a pattern of low-frequency activity (1 to 3.5
Hz) occurring at specific times and scalp locations is a reliable
correlate of human signal detection performance.
From: Aggelos Katsaggelos <aggk@delta.eecs.nwu.edu>
Subject: CFP: Visual Communications and Image Processing '94 (VCIP '94)
ANNOUNCEMENT AND CALL FOR PAPERS
Visual Communications and Image Processing '94 (VCIP '94)
September 25-28, 1994, Bismark Hotel, Chicago, IL, USA
Cooperating Organizations:
IEEE Circuits and Systems Society
European Association for Signal Processing (EURASIP)
Conference Chair:
Aggelos K. Katsaggelos, Northwestern Univ.
Europe Liaison:
Prof. Jan Biemond, Delft Univ. of Technology,
Department of Electrical Engineering, Biemond@et.tudelft.nl
Asia Liaison:
Dr. Toshi Koga, NEC Corp, Second Transmission Division
koga@trdv1.trd.tmg.nec.co.jp
Visual communications and image processing have become engineering
areas that attract interdisciplinary research interest and lead to
significant developments for technology and science. This conference
is designed as a forum for presenting important research results in
these fields. Original and unpublished material of novel techniques and
new developments is solicited on the following and related topics:
A. Visual Communications
still image coding -- video sequence coding -- model-based image coding
very low bit rate coding -- motion estimation for video coding
TV, HDTV and super HDTV -- packet video -- picture archiving and
communications systems -- parallel processing hardware -- special topics
B. Image Processing and Analysis
image filtering, enhancement, and restoration -- feature extraction
image segmentation -- object recognition -- biomedical image processing
VLSI implementation and system architectures -- morphological image
processing -- fractals and wavelets -- special topics
C. Image Sequence Analysis
motion analysis -- filtering, restoration, interpolation -- multiscale
image sequence analysis -- computer vision topics in image sequence
analysis -- special topics
Best Student Paper Awards
At this meeting, two awards will be presented to the best papers
submitted by students. To qualify for these awards, each consisting of
a cash prize and a plaque, the candidate must be the principal author.
A letter from the student's adviser stating that the major work was
done by the student must accompany the final manuscript in order to be
considered by the award committee.
SPIE gratefully acknowledges Lockheed Corp. and NEC Corp. for
their generous sponsorship of these awards.
Young Investigator Award
An award will be given for the best paper submitted by a young
researcher who has graduated within five years from the date of the
meeting. A letter from the researcher requesting that his/her paper be
considered must be submitted with the final manuscript, certifying
he/she meets the qualifications for the award.
SPIE gratefully acknowledges Industrial Technology Research Institute
(ITRI) for generously sponsoring this award.
SUBMISSION OF SUMMARIES:
Summary Due Date: 24 January 1994* Manuscript Due Date: 20June 1994+
*Late submissions may be considered, subject to program time
availability and chairs' approval. However, due to the significant
number of submissions, chairs and committee will give first preference
to summaries sent by the due date.
+Proceedings of this conference will be published before the meeting
and available at the symposium. Manuscript due date must be strictly
observed.
Your summary should include the following:
1. SUMMARY TITLE
2. AUTHOR LISTING (principal author first)
Full names and affiliations as they will appear in the program.
3. CORRESPONDENCE FOR EACH AUTHOR
Mailing address, telephone, telefax, e-mail address.
4. SUBMIT TO: Visual Communications and Image Processing '94
5. INDICATE: Letter of topic (A, B, or C) and subtopic
6. PRESENTATION
Please indicate your preference for either "Oral Presentation" or
"Poster Presentation".
7. SUMMARY TEXT 500 words.
8. BRIEF BIOGRAPHY (principal author only) 50 to 100 words.
To be considered for acceptance, you may either:
* send summary via electronic mail to Internet abstracts@mom.spie.org
(ASCII format);
* or fax one copy to SPIE at 206/647-1445;
* or mail four copies to SPIE to:
Visual Communications and Image Processing '94
SPIE, P.O. Box 10, Bellingham, WA 98227-0010
Shipping address: SPIE, 1000 20th Street, Bellingham, WA 98225
SPIE contact: Rosa Cays, Coordinator, Technical Programs and Education
Phone: 206/676-3290; Telex 46-7053
Note from the editor: this announcement was shortened for inclusion
in the digest, please contact the organizers for the long version.
From: Daniel K R McCormack <Daniel.K.R.McCormack@cm.cf.ac.uk>
Subject: Looking for code in Daubechies book
Hello,
I want to plot the scaling function for D4, D12 and D20 wavelets.
There is an algorithm given in Chp 6 section 5 of I. Daubechies book
"Ten Lectures on Wavelets" for calculating an approximation to the scaling
function. I would like to know if anyone has coded this
algorithm or any other similar algorithm. Any reference to this code
would be much appreciated.
Alternatively if someone could reference or give me the values
of the function evaluation for D4, D12 and D20 it would save me some time
and effort.
Many Thanks
Daniel McCormack.
From: aldroubi@helix.nih.gov
Subject: Preprint available: Portrait of frames
Preprint of the paper below is available: if interested, send your postal
address by email to aldroubi@helix.nih.gov
Portrait of frames
(To appear in the Proc. Amer. Math. Soc.)
By Akram Aldroubi
We introduce two methods for generating frames of a Hilbert space
$\cal H$. The first method uses bounded operators on $\cal H$. The
other method uses bounded linear operators on $l_2$ to generate
frames of $\cal H$. We characterize all the mappings that transform
frames into other frames. We also show how to construct all frames
of a given Hilbert space $\cal H$, starting from any given one. We
illustrate the results by giving some examples from multiresolution
and wavelet theory.
From: Ravindra Chandrashekhar <rachan@utdallas.edu>
Subject: Question: Wavelets for Image Compression
Dear Wavelet Experts,
I started researching Wavelets (Discrete Wavelet Transform, DWT) this
semester and have read several papers. There seems to be an abundance
of papers, books, and mathematical reviews. I have also reviewed some
of the publication listings under the wavelets directory on this network.
I am trying to do research on image compression. I have read articles by
O. Rioul, Shensa et al and am in the process of reviewing "Image Compression
Through Wavlet Transform Coding" by DeVore, Jawerth, and Lucier.
1) So far, I have obtained articles through my own research. Any
suggestions on "Must Read Articles" related to image compression that
I may have missed.
2) Can any one suggest articles related to image compression that have code
examples or sources of code? I reviewed the DWT in the Numerical Reciepies
Book. But I am open to suggestions on other code examples and articles.
I would be very happy to receive any suggestions.
My email address is rachan@utdallas.edu
Thanks!
Sincerely,
Ravindra Chandrashekhar
From: Peter Schröder <ps@princeton.edu>
Subject: Computing line integrals of a 3D function described in wavelets?
Question: How to compute line integrals of a 3D function described in
wavelets?
Consider a scalar function f(x,y,z)=f(\vex{x}) (sufficiently smooth)
defined on a lattice [0..127]^3. Assume further that we have already
transformed it into the wavelet basis. We would now like to compute
\exp(-\int_{t_0}^{t_1} f(\vec{x}_0+s(\vec{x}_1-\vec{x}_0)) ds)
or in words, the exponential of -f() integrated along some ray from a
starting point to some endpoint in the original volume. Why? Well, if f()
represents the total scattering cross section of some material the above
integral will give us optical depth in the given medium along a line of
sight. This is only a subproblem of a larger problem.
Now, obviously this is easy in the 1D case since path integration is
trivial but things become a lot less straightforward in higher D. In
particular there is no hope of somehow precomputing the above \exp()
integrals for all possible choices of (x,y,z)_{t_0} (x,y,z)_{t_1} since
that is a 6 dimensional object and even if we can do the wavelet transform
quickly in one dimension the cost of doing it along all dimensions is going
to be totally overwhelming. Knowing this it seems we need some lazy scheme
to evaluate it, i.e. somehow take advantage of the sparse structure of the
original wavelet representation of f() and directly compute the above
function for a given (t_0 t_1). Of course the \exp() should help *a lot* in
this task as well since the tails will contribute very little and we should
be able to take advantage of this in the multi resolution framework.
Presumably this would be helped considerably if we knew something about
objects of the form
\exp(-\int_{t_0}^{t_1} \phi(\vex{x}_0+s(\vec{x}_1-\vec{x}_0)))
and similarly for \psi. Or in words, \exp()s of path integrals through
wavelet bases (product bases) in 3D. Does anybody have any experience with
this? References? War stories?
Thanks for any help you might have!
Peter (ps@princeton.edu)
From: Bruno Torresani <torresan@marcptnx1.univ-mrs.fr>
Subject: Papers available:
The following papers are available by anonymous ftp at
cpt.univ-mrs.fr (139.124.7.5)
login as "anonymous" and enter your e-mail address as password.
path:
/pub/preprints/93/wavelets
The following preprints are available
---
92-P.2811.tar.Z (Plain TeX)
\centerline{\bf N-DIMENSIONAL AFFINE WEYL-HEISENBERG WAVELETS}
\centerline{\bf C. Kalisa, B.Torresani}
{\bf Abstract} : Multidimensional coherent states systems
generated by translations, modulations, rotations and dilations are
described. Starting from unitary irreducible representations of
the $n$-dimen\-sional Weyl-Heisenberg group, which are not
square-integrable,
one is led to consider systems of coherent states labeled by the
elements of
quotients of the original group. Such systems can yield a resolution
of the
identity, and then be used as alternatives to usual wavelet of
windowed
Fourier analysis. When the quotient space is the phase space of the
representation,
different embeddings of it into the group provide different
descriptions of the phase space.
September 1992;
to appear in "Annales de l'institut Henri Poincar\'e, Physique
Th\'eorique".
---
93-P.2910.tar.Z (latex, with postscript figures)
\centerline{\bigbf PYRAMIDAL ALGORITHMS FOR}
\centerline{\bigbf LITTLEWOOD-PALEY DECOMPOSITIONS}
\centerline{M.A. Muschietti, B. Torr\'esani }
\noindent{\bf Abstract} : It is well known that with any usual
multiresolution
analysis of $L^2({\bf R})$ is associated a pyramidal algorithm for
the
computation of the corresponding wavelet coefficients.
It is shown that an approximate pyramidal
algorithm may be associated with more general Littlewood-Paley
decompositions.
Accuracy estimates are provided for such approximate algorithms.
Finally, some
explicit examples are studied.
May 1993; To appear in SIAM J. Math. An.
---
93-P.2932.tar.Z (latex, with postscript figures)
\centerline{\bf WAVELETS ON DISCRETE FIELDS}
\centerline{\bf K.Flornes, A.Grossmann, M.Holschneider,
B.Torr\'esani}
\begin{abstract}
An arithmetic version of continuous wavelet analysis is described.
Starting
from a square-integrable representation of the affine group of $\Z_p$
(or $\Z$)
it is shown how wavelet decompositions of $\ell^2(\Z_p)$ can be
obtained.
Moreover, a redefinition of the dilation operator on $\ell^2(\Z_p)$
directly yields an algorithmic structure similar to that appearing
with
multiresolution analyses.
\end{abstract}
July 1993; To appear at Applied and Computational Harmonic Analysis.
---
93-P.2878.tar.Z latex (with postscript figures)
\centerline{\bigbf PHASE SPACE DECOMPOSITIONS:}
\centerline{\bigbf Local Fourier analysis on spheres}
\centerline{Bruno Torresani}
\noindent{\bf Abstract} : Continuous wavelet analysis
and Gabor analysis have proven
to be very useful tools for the analysis of signals in which local
frequencies can be extracted. Both techniques can be described
in the same footing using the theory of square-integrable group
representations or derived theories. It is shown here how the
same kind of techniques can be developed to construct
phase-space representation theormes for functions defined
on homogeneous spaces. As examples, the cases of the circle
and the spheres are carried in details, and some explicit expressions
are given in the one and two-dimensional cases.
March 1993;
Submitted to Applied and Computational Harmonic Analysis.
---
93-P.2870.tar.Z (Plain TeX, with postscript figures)
\centerline{\bf LOCAL FREQUENCY ANALYSIS WITH}
\centerline{\bf TWO-DIMENSIONAL WAVELET TRANSFORM}
\centerline{\bf Caroline Gonnet \hskip2cm Bruno Torresani}
{\bf Abstract} : An algorithm for the characterization of local
frequencies
in two-dimensional signals is described. The algorithm generalizes to
the 2D
situation a method previously proposed in the 1D context, based on
the analysis
of the phase of the continuous wavelet transform. It uses families of
wavelets
generated from a unique one by shifts, dilations and rotations.
January 1993; revised April 1993.To appear in Signal Processing
From: gs@BOURBAKI.MIT.EDU
Subject: Short course on wavelets filter banks and applications
SHORT COURSE on WAVELETS FILTER BANKS and APPLICATIONS
at Wellesley College near Boston: June 24-27 1994
organized and taught by Gilbert Strang and Truong Nguyen , M I T
This is the first announcement of a short course on filter banks,
multirate signal processing, and wavelets. By sending an email
message to Gilbert Strang at MIT (( gs@math.mit.edu ))
you will receive the second announcement directly. The 4-day course
will be given over the last weekend in June, 1994, to reduce air fares
and allow for sightseeing in Boston and New England. Wellesley College
has the most beautiful campus in this area ( 14 miles west of Boston ).
Room and board at the College is available to participants at reasonable cost.
The workshop course will be taught by Gilbert Strang and Truong Nguyen of MIT.
The topics will include
Digital Filters: Analysis and Design
Matrix Analysis: Toeplitz Matrices and Circulants
Multirate Signal Processing: Filtering, Decimation, Polyphase
Wavelet Transform: Pyramid Implementation
Daubechies Wavelets, Biorthogonal Wavelets, Vector Wavelets
Compression, Transient Detection, Radar Processing
Non-Destructive Evaluation, Digital Communications
Perfect Reconstruction: Orthonormal, Cosine-modulation
Spectral Factorization, Transmultiplexers
Quantization effects
This workshop course will NOT assume an extended background in
filter banks and wavelets. It is developed from an MIT graduate course
emphasizing applications of the theory. Participants are invited to
suggest problems and possible applications for discussion.
An email message to gs@math.mit.edu mentioning "Workshop Course" will
bring further information early in 1994.
<webmaster@wavelet.org>