Copy Link
Add to Bookmark
Report
VISION-LIST Digest Volume 14 Issue 26
VISION-LIST Digest Mon Jul 10 11:00:52 PDT 95 Volume 14 : Issue 26
- ***** The Vision List host is TELEOS.COM *****
- Send submissions to Vision-List@TELEOS.COM
- Vision List Digest available via COMP.AI.VISION newsgroup
- If you don't have access to COMP.AI.VISION, request list
membership to Vision-List-Request@TELEOS.COM
- Access Vision List Archives via anonymous ftp to TELEOS.COM
Today's Topics:
Edge thresholding code available
Skeletonization Software (Info on Downloading)
Announce: ROBOSOFT WWW server
Raid for real-time image capture?
WWLab
Looking for Tangent Distance Implemented in C or C++
Q: automatic recognition of number plates, anyone?
Animated eigenmode datasets
Research positions in Image Understanding at Siemens Corporate Research
Ph.D. Fellowships available for U.S. citizens and permanent residents
Report available
CFP: 12th Israeli Symposium on AI, Computer Vision and NN
Machine Vision Symposium
Autumn School in Connectionism and Neural Networks
Short Course on Image Compression: 9/13-9/15 Portland, Oregon
----------------------------------------------------------------------
Date: Fri, 30 Jun 1995 10:09:49 +0200
From: paul.rosin@jrc.it (Paul Rosin)
Subject: Edge thresholding code available
In the paper "Dynamic threshold determination by local and global edge
evaluation", Svetha Venkatesh & Paul L. Rosin, Graphical Models &
Image Processing, Vol. 75, No. 2, pp. 146-160, 1995, we describe a
method for performing automatic thresholding of thin connected edges.
The code is now available from the Vision List Archives at:
ftp://ftp.teleos.com/VISION-LIST-ARCHIVE/SHAREWARE/CODE/EDGE-THRESHOLDING/
Some further work is described in the Techical note No. I.95.58, "Edges:
Saliency measures and automatic thresholding", Paul L. Rosin,
Institute of Remote Sensing Applications, Joint Research Centre, which
is also available from: ftp://gamma.jrc.it/pub/paul/edges2.ps.Z
Paul Rosin
Institute of Remote Sensing Applications
Joint Research Centre
Ispra (VA), I-21020
Italy
email: paul.rosin@jrc.it
------------------------------
Date: 7 Jul 1995 16:22:56 GMT
From: rlo@arcadia.harvard.edu (Robert Ogniewicz)
Organization: Harvard Robotics Laboratory
Subject: Skeletonization Software (Info on Downloading)
Hi Everybody,
This is an update on my first posting (June 28):
> I have prepared a first public domain release of my
> 2-D skeletonization algorithms. The package can be fetched
> through access to my WWW homepage, URL =
> http://hrl.harvard.edu/people/postdocs/rlo.html or from
> our FTP server ftp.hrl.harvard.edu (directory pub/skeletons,
> file MAT-1.0alpha.tar.Z).
> etc...
During the last days, many people have attempted to retrieve software
and postscript copies of my publications through WWW. Unfortunately,
some persons reported that they were not able to copy neither papers
nor software. While it is possible to directly download to local disk
(in Mosaic, select "load to local disk" from menu "options"), I
understand that the current setup of my homepage was not optimal.
Currently, all publications plus software are also accessible from
ftp.hrl.harvard.edu (directory pub/skeletons). An INDEX file
describes the contents of this directory. All corresponding hyperlinks
on my homepage are now pointing at ftp://hrl.harvard.edu. Downloading
through access to my homepage should be straightforward.
Regards,
Robert Ogniewicz
------------------------------
Date: 5 Jul 1995 15:02:16 GMT
From: buffa@essi.fr (Michel.BUFFA)
Organization: Universite de Nice
Subject: Announce: ROBOSOFT WWW server
Since 1985, ROBOSOFT is the European leader in the field of mobile
robotics. ROBOSOFT can provide OEMS, integrators and end-users with
all products and services in mobile robots and intelligent vehicle
design.
Robosoft is now on the Web:
http://www.essi.fr/robosoft/
Michel Buffa: Ecole Superieure des Sciences Informatiques (ESSI)
650 Routes des Colles, BP 145
06903 Sophia-Antipolis CEDEX
Internet: buffa@essi.fr
Voice phone: (33) 92-96-51-50
Fax: (33) 92-96-51-55
WWW home page:
http://www.essi.fr/~buffa
Check my WWW video game pages:
http://www.essi.fr/~buffa/videogames.html
------------------------------
Date: Thu, 6 Jul 95 14:33:20 -0500
From: seitz@cs.wisc.edu (Steve Seitz)
Subject: Raid for real-time image capture?
We are looking into using RAID arrays for real-time image capture in vision
applications. The goal would be to capture and store high-quality
uncompressed images at 30 frames per second. Ideally the images would be
roughly 640x480 >=8 bit monochrome or color but these are not hard
constraints. There seem to be a variety of capture cards on the market for
all platforms (we are using Sun Sparc 10's) but not much in the way of
real-time STORAGE media.
Has anyone found a satisfactory solution for real-time image capture using
RAID or other fast media? We are currently looking into Abekas as a
possibility, but their products may be out of our price range.
Please respond via email to seitz@cs.wisc.edu.
Thanks!
Steve Seitz
Dept. C.S.
University of Wisconsin--Madison
(608) 262-6619
------------------------------
Date: Sat, 01 Jul 1995 00:35:41 +0200
From: Frank Mangin <Frank.Mangin@sophia.inria.fr>
Subject: WWLab
An html version of this document is available in
http://www.inria.fr/pastis/wwl/welcome.html
Join the World Wide Lab project!
NEW! A demo server is now available
WHAT IS IT?
Tools allowing inter-project cooperation and evaluation of concurrent works
have long been identified as a crucial need in the vision research field. The
purpose of the World Wide Lab project is to provide such an easy to use, widely
available tool. The key idea is to allow researchers to make the execution of
their programs available to the research community on the World Wide Web.
WHO'S CONCERNED?
Providers:
you just published a paper describing the algorithm of the year: give
others a chance to test it with their own data and parameters by
registering the corresponding program at the WWLab. This doesn't require
to make the program itself available if you don't want to!
Clients:
tired of reading about algorithms you never see at real work? Tired of
wondering how this new approach compares to your's? At the WWLab, you will
be able to choose a program, feed it with your data, and get back the
result.
HOW WILL IT WORK?
The WWLab will be accessed through a World Wide Web form: once the client has
specified the program to execute, along with the data and parameters to use,
the request is acknowledged and recorded on the WWLab server for later
processing. When the result is available, the client is notified by email that
he can retrieve it.
Researchers willing to make a program available will have to provide the code
(see "Keypoints" below) of the program, input and output data descriptions, and
related scientific publications or references. All submitted information,
including code, remains the property of the author, who can at any time ask
that it be removed from the WWLab. The WWLab will home the executable, and run
it on its own dedicated machines as client requests are processed.
KEYPOINTS
** In order to encurage submissions, the required effort to make a program
available will be maintained as low as possible: no source code rewriting, no
new "universal" language or huge C++ class library...
** Each provider keeps full control over the material he submitted to the
WWLab: programs will be kept on dedicated machines, in executable form only,
they won't be distributed in any way, and will be removed at any time on the
author's request.** This is a non-profit project whose only purpose is
to provide a
valuable service to the research community.
HOW YOU CAN HELP
This project is under construction, we need your help! Any form of support will
be welcome, e.g.:
** you think this is a valuable idea?
Send us a mail explaining why you think so, include the responsibilities
you may have in the research field: your support will help in convincing
others, in finding fundings...
** you would be ready to make some of your programs available?
Providers are at the heart of the project, so let us know - this won't be
taken as an engagement of your part.
** you would like to participate in the project:
all the better! We will need software engineering, mirror sites, computing
platforms to process requests...
HOW TO CONTACT US
** On the Web: this page and all WWLab related information can be
retrieved in
http://www.inria.fr/pastis/wwl
Check there for future evolution of the project.
** By Email:
wwlproject@sophia.inria.fr
** By ground mail:
Frank Mangin
Projet WWLab
INRIA
2004 route des lucioles
BP 93 - 06902 Sophia-Antipolis Cedex
FRANCE
Frank Mangin
Inria Phone #93 65 78 66
2004 route des Lucioles Email: mangin@mirsa.inria.fr
06560 Valbonne Cedex
France
------------------------------
Date: Thu, 29 Jun 1995 14:26:39 -0700
From: Rushdy@SSTCX1.Lanl.GOV (Q.Rushdy Ahmad)
Organization: LANL
Subject: Looking for Tangent Distance Implemented in C or C++
I am working on a Nueral Network Library for DOE. We have Back
Propigation Nets, Self organizing Maps, and classical distance
measures for classifing image data. We are looking for an
implementation of the Tangent Distance, which is is an invarient
distance measure for (slight rotation, scaling, etc.) for two
dimensional matricies (images).
We found the distance Alg. in "Efficent Pattern Recognition Using New
Transformation Distance" by Patrice Simard, Yann Le Cun, and John
Denker of Bell Labs. They implemented the Alg. for use in character
recognition and pattern recognition, we are interested in using the
algo. for getting better image distances from proto- types in Image
Analysis and LandStat analysis.
I have contacted Bell Labs and their implemntation is proprietary.
I was hoping that some one else had implemented this in a library
of matrix functions.
Please give me pointers to where this sort of function could be found???
thankyou for any help in advance,
Jim McNeil
LANL
pjm@cotopaxi.esa.lanl.gov
------------------------------
Date: Wed, 05 Jul 1995 13:22:17 +0200
From: arnoudv@ped.kun.nl (Arnoud Verdwaald)
Organization: Special Education, Nijmegen University
Subject: Q: automatic recognition of number plates, anyone?
Hi,
is there anyone out there who could give me addresses (pref. fax
numbers and contact persons) of companies or others that have
developed a system for automatic recognition of number plates of
vehicles? Key words: REGISTRATION of TRAFFIC VIOLATIONS, especially
SPEEDING.
Please respond by e-mail, since I don't follow this newsgroup (if this
posting doesn't belong in this newsgroup, I apologize; I do this for a
friend, this is not my cup of tea).
Thanks a lot!
Arnoud (e-mail: A.Verdwaald@ped.kun.nl)
------------------------------
Date: 30 Jun 1995 17:14:04 GMT
From: Giovanni De Gasperis <degas@solid.mda.uth.tmc.edu>
Organization: Univ. of Texas M.D. Anderson Cancer Center
Subject: (no subject)
Dear netters,
I'm new in the field of computer vision, so please, be patient
with my naive questions.
I'going to investigate a good and robust algorithm, or a combination
of hardware & software, with similar features:
- real time processing (on a 64x64x8bit/pixel region of interest)
- capable of tracking a rotating object
- make measurements of translational and rotational speed
of the object
- distinguish an object by an other very similar in shape
during all the measurements.
Is it science fiction, or do you think I cand find something like
this?
Thank you in advance
Giovanni De Gasperis Visiting Scientist
degas@solid.mda.uth.tmc.edu
Experimental Pathology Department Box 089
University of Texas - M. D. Anderson Cancer Center
------------------------------
Date: 10 Jul 1995 17:08:48 GMT
From: mhs@eng.cam.ac.uk (M.H. Syn)
Organization: Cambridge University Engineering Department, UK
Subject: Animated eigenmode datasets
An animated demonstration of eigenmode data from the technical report:
CUED/F-INFENG/TR211
"FEM Eigenmodes as Shape Features"
Michael Syn and Richard Prager
is available for the following platforms:
Linux 1.2
SunOS 4.1.3 / Solaris 2.3
IRIX 5.3
HP-UX 9.01
Anonymous ftp from svr-ftp.eng.cam.ac.uk:
/pub/data/geomdemo.README
/pub/data/geomdemo.linux.tar.gz
/pub/data/geomdemo.sun.tar.gz
/pub/data/geomdemo.sgi.tar.gz
/pub/data/geomdemo.hp.tar.gz
/pub/reports/syn_tr211.ps.gz
mhs@eng.cam.ac.uk
Speech, Vision and Robotics Group, Cambridge University Engineering Dept.
------------------------------
Date: Wed, 5 Jul 1995 15:52:01 -0400
From: Jian-Zhong Qian <qian@scr.siemens.com>
Subject: Research positions in Image Understanding at Siemens Corporate Research
At Siemens, a $50 billion global company, we recognize the value of R&D.
That's why we invented approximately $5 billion in worldwide R&D
efforts last year. Siemens Corporate Research in Princeton NJ
is an integral part of centralized corporate R&D with partners in
Munich and Erlangen (Germany). At our 23-acre complex, we perform
research and basic development in imaging and visualization,
multi-media technology, software engineering, and learning systems.
Siemens Corporate Research, Inc. is currently looking for highly
talented researchers in the areas of imaging and visualization,
specifically, in image processing, analysis, and understanding.
The successful candidate will have a Ph.D. or MS degree in CS/EE,
with excellent abilities in research and publications,
innovative thinking, real world problem solving, and fast prototyping.
Strong expertise in implementation and programming is a plus for
candidates with MS degrees.
We offer competitive salaries and comprehensive benefits.
For immediate consideration, please send your resume to:
Human Resources, Siemens
Corporate Research, Inc.,
755 College Road East,
Princeton, NJ 08540.
Fax (609) 734-3605.
or
Email to: qian@scr.siemens.com (ASCII or Postscript).
Equal Opportunity Employer.
------------------------------
Date: Fri, 30 Jun 1995 09:32:55 +0600
From: canning@alonzo.cs.uoknor.edu (John M. Canning)
Subject: Ph.D. Fellowships available for U.S. citizens and permanent residents
The University of Oklahoma School of Computer Science has several GAANN
fellowships available on a competitive basis to prospective PhD students.
Prof. John Canning (jcanning@uoknor.edu) and Prof. Deborah Trytten
(trytten@mailhost.ecn.uoknor.edu) work in the area of computer vision and
encourage students planning on seeking a Ph.D. to apply to our program. The
fellowship announcement follows:
The School of Computer Science is Seeking GAANN Fellows to
pursue PhDs in Computer Science
Up to $14,400 annual stipend (depending on need)
Full coverage of tuition, fees, and required textbooks
Additional funds for travel, supplies, and special equipment
Dedicated Unix workstation for each Fellow
Fellows must
Be excellent students, qualified for the PhD program
Intend to pursue academic careers
Be United States citizens or permanent residents
or intend to become permanent residents of the United States
or be permanent residents of the Trust Territory of the Pacific Islands
Make steady progress and complete their degrees within five years
The University of Oklahoma (OU) is
a major, national research university serving the educational,
cultural and economic needs of the state, region and nation. OU
enrols more than 25,000 students on campuses in Norman, Oklahoma City
and Tulsa ad has approximately 1,500 full-time faculty members.
Apply to
GAANN Fellowship Director
School of Computer Science
University of Oklahoma phone 405-325-4042
200 Felgar Street - EL114 fax 405-325-4044
Norman OK 73019 email cs@uoknor.edu
The University of Oklahoma is an equal opportunity/affirmative action employer.
Women and minorities are especially encouraged to apply.
GAANN (Graduate Assistance in Areas of National Need) is a program of the
United States Department of Education
UNIX is a registered trademark of UNIX System Laboratories
------------------------------
Date: Tue, 4 Jul 1995 12:22:13 +0200
From: ronse@dpt-info.u-strasbg.fr
Subject: Report available
The following report, whose abstract is given below, is available by
anonymous ftp. It consists in two Postcript files, one for the text, and
the other for the figures; these files are compressed. In order to obtain
them, you do:
ftp dpt-info.u-strasbg.fr
Name: anonymous
Password: <your_e-mail_address>
ftp> cd pub/recherche/Vision
ftp> get rap95-07.ps.Z
ftp> get rap95-07_fig.ps.Z
ftp> bye
uncompress rap95-07.ps
uncompress rap95-07_fig.ps
and then print them.
If you want a printed version instead of Postcript files, send me a request
by e-mail, with your complete postal address (including zip code and
country). You must wait for printing delays.
A lattice-theoretical morphological view on template extraction in images
Christian Ronse
LSIIT - URA 1871
Universite Louis Pasteur
UFR de Mathematique et Informatique
7 rue Rene Descartes
F-67000 Strasbourg, France
e-mail: ronse@dpt-info.u-strasbg.fr
ABSTRACT:
The extraction of copies of a template appearing in a figure can be defined
as the composition of a hit-or-miss transform by a pair of structuring
elements, followed by a dilation by the first structuring element. On the
other hand, the operation of template extraction can be abstractly
characterized by the three requirements of anti-extensivity, idempotence,
and a new property that we call ``overcondensation''; we designate by
``open-overcondensation'' an operator having these three properties.
Examples include digital contour extraction, or the above-mentioned
composition of a hit-or-miss transform by a dilation. Decompositions
formulas of open-overcondensations in terms of such operators are given;
they parallel the well-known decomposition formulas for openings.
Generalisations of ``rank-max'' openings and inf-overfilters are also
given. These results hold not only for sets or grey-level images, they are
valid in the framework of complete lattices.
KEY WORDS:
template extraction, mathematical morphology, hit-or-miss transform,
idempotence, overcondensation
Christian Ronse ronse@dpt-info.u-strasbg.fr
LSIIT - URA 1871
Universite Louis Pasteur
UFR de Mathematique et Informatique
7 rue Rene Descartes Tel. (33) 88.41.66.38
F-67000 Strasbourg Fax. (33) 88.61.90.69
------------------------------
Date: Fri, 30 Jun 95 13:15:14 IDT
From: mic@csa.cs.Technion.AC.IL (Michael Lindenbaum)
Subject: CFP: 12th Israeli Symposium on AI, Computer Vision and NN
Announcement and Call For Papers
12th Israeli Symposium on
Artificial Intelligence,
Computer Vision and Neural Networks
Tel Aviv University, Tel Aviv, February 4-5, 1996
The purpose of the symposium is to bring together researchers and
practitioners from Israel and abroad interested in the areas of
Artificial Intelligence, Computer Vision, and Neural Networks, and to
promote interaction between them. The program will include
contributed as well as invited lectures and possibly some tutorials.
All lectures will be given in English.
Papers are solicited addressing all aspects of AI, Computer Vision and
Neural Networks. Novel contributions in preliminary stages are
especially encouraged but significant work which was already presented
recently will also be considered.
The symposium is intended to be more informal than the previous ones.
The proceedings, including summaries of the contributed and invited
talks, will be organized as a technical report and distributed during
the symposium. No copyright will be required. To minimize costs, we
intend to organize it in a university.
Authors should submit an extended abstracts of their presentation in
English to be received by September 1st 1995.
Submissions must be limited to four
pages including title and bibliography. Submitted contributions will be
refereed by the program committee. Authors will be notified of
acceptance by November 1st, 1995. A final abstract to be
included in the proceedings is due January 10, 1996.
For receiving updated information on the symposium, please send a
message to Yvonne Sagi, (yvonne@cs.technion.ac.il) including your name,
affiliation, e-mail, fax number and phone number.
Submitted summaries should be send to:
Yvonne Sagi
Computer Science Department
Technion, Israel Institute of Technology
Haifa, 32000
Israel
Program chairmen: Dan Geiger, Micha Lindenbaum and Ron Meir
Dan Geiger (Artificial Intelligence)
e-mail: dang@cs.technion.ac.il
Phone: 972-4-294265
Micha Lindenbaum (Computer Vision)
e-mail: mic@cs.technion.ac.il
Phone: 972-4-294331
Ron Meir (Neural Networks)
e-mail: rmeir@ee.technion.ac.il
Phone: 972-4-294658
------------------------------
Date: Fri, 7 Jul 1995 14:26:07 +0100
From: "Harald Rihl" <HRIHL@novell.pr-steyr.ac.at>
Organization: profactor steyr
Subject: Machine Vision Symposium
What: International Symposium
Title: Machine Vision in the industrial practice
Date: 5th. and 6th. of September 1995
Location: Steyr, Austria
Further information:
URL: http://www.pr-steyr.ac.at/bv-symposium/
PROFACTOR Produktionsforschungs GmbH.
VPTOE Vereinigung z. Foerderung d. Produktionstechnologien in Oesterreich
Dipl.-Ing. Harald Rihl
Wehrgrabengasse 5
A-4400 Steyr, Oesterreich
Tel.Nr.: +43/07252-884-232
Fax Nr.: +43/07252-884-244
E-mail : Harald.Rihl@pr-steyr.ac.at
------------------------------
Date: 4 Jul 1995 10:02:01 GMT
From: paass@gmd.de (Gerhard Paass)
Organization: GMD, Sankt Augustin, Germany.
Subject: Autumn School in Connectionism and Neural Networks
CALL FOR PARTICIPATION
=================================================================
= = = H e K o N N 9 5 = = =
Autumn School in
C o n n e c t i o n i s m and N e u r a l N e t w o r k s
October 2-6, 1995
Muenster, Germany
Conference Language: German
A comprehensive description of the Autumn School together with
abstracts of the courses can be found at the following
addresses:
WWW: http://borneo.gmd.de/~hekonn
anonymous FTP: ftp.gmd.de
directory: Learning/neural/hekonn95
= = = O V E R V I E W = = =
Artificial neural networks (ANN's) have in recent years been
discussed in many diverse areas, ranging from the modelling
of learning in the cortex to the control of industrial
processes. The goal of the Autumn School in Connectionism
and Neural Networks is to give a comprehensive introduction
to conectionism and artificial neural networks and to give
an overview of the current state of the art.
Courses will be offered in five thematic tracks. (The
conference language is German.)
The FOUNDATION track will introduce basic concepts (A. Zell,
Univ. Stuttgart), as well as present lectures on information
processing in biological neural systems (G. Palm, Univ. Ulm),
on the relationship between ANN's and fuzzy logic (R. Kruse,
Univ. Braunschweig), and on genetic algorithms (S. Vogel,
Univ. Cologne).
The THEORY track is devoted to the properties of ANN's as
abstract learning algorithms. Courses are offered on
approximation properties of ANN's (K. Hornik, Univ. Vienna),
the algorithmic complexity of learning procedures
(M. Schmitt, TU Graz), prediction uncertainty and model
selection (G. Paass, GMD St. Augustin), and "neural"
solutions of optimization problems (J. Buhmann, Univ. Bonn).
This year, special emphasis will be put on APPLICATIONS of
ANN's to real-world problems. This track covers courses on
vision (H.Bischof, TU Vienna), character recognition
(J. Schuermann, Daimler Benz Ulm), speech recognition
(R. Rojas, FU Berlin), industrial applications
(B. Schuermann, Siemens Munich), robotics (K.Moeller, Univ.
Bonn), and hardware for ANN's (U. Rueckert, TU
Hamburg-Harburg).
In the track on SYMBOLIC CONNECTIONISM, there will be
courses on: knowledge processing with ANN's (F. Kurfess, New
Jersey IT), hybrid systems in natural language processing
(S. Wermter, Univ. Hamburg), connectionist aspects of
natural language processing (U. Schade, Univ. Bielefeld),
and procedures for extracting rules from ANN's (J. Diederich,
QUT Brisbane).
In the section on COGNITIVE MODELLING, we have courses
on representation and cognitive models (G. Dorffner,
Univ. Vienna), aspects of cognitive psychology
(R. Mangold-Allwinn, Univ. Saarbruecken), self-organizing
ANN's in the visual system (C. v.d. Malsburg, Univ. Bochum),
and information processing in the visual cortex
(J.L. v. Hemmen, TU Munich).
In addition, there will be courses on PROGRAMMING and
SIMULATORS. Participants will have the opportunity to work
with the SESAME system (J. Kindermann, GMD St.Augustin) and
the SNNS simulator (A.Zell, Univ. Stuttgart).
------------------------------
Date: 7 Jul 1995 11:29:25 -0700
From: fli@ee.pdx.edu (Fu Li)
Organization: Portland State University, Portland, OR
Subject: Short Course on Image Compression: 9/13-9/15 Portland, Oregon
Three-Day Intensive Course on
Digital Image and Multimedia Compression:
Fundamentals and International Standards
Wednesday, September 13 - Friday, September 15, 1995
For more information, please see our WWW homepage at
http://www.ee.pdx.edu/short_courses/image_compression/
Seats are very limited, early registration is encouraged!
About the Course:
Digital cameras and Photo CD digital negatives, digital television broadcast
and movies stored on CDROM that can be played back on reasonably priced
multimedia PCs, are already a reality. Due to the huge amounts of data
associated with digital images and video, their efficient storage and
transmission poses a challenging problem. Consequently, image and video
compression plays an enabling role for almost all consumer, commercial and
scientific applications.
Digital image and video compression is a current focus of both research
and international standardization. The recently developed standards such
as JPEG, JBIG, H.261, MPEG-1, and MPEG-2 reflect the state of the art
algorithms and are important in facilitating interoperability among various
imaging systems as well as wide-spread, cost-effective deployment of the
technology.
This course provides the audience with the necessary foundation: the
principles of information theory, motion estimation, and motion compensated
preprocessing. A detailed discussion of compression algorithms for both
still images and video is presented on the basis of this foundation, enabling
a solid understanding of the state of the art, standards and future directions.
Upon completion of the course, the students will be equipped with the necessary
background in information theory, motion estimation and compensation, and a
solid understanding of the basic principles and standards of digital image
and video compression.
Presenters:
Majid Rabbani and M. Ibrahim Sezan
Image Processing Laboratory
Eastman Kodak Company
Organizers:
Fu Li and Rolf Schaumann
Portland State University
With Cooperation by
Oregon Center for Advanced Technology Education
The course starts with a broad set of product and application examples that
establish the need for image compression in various digital imaging systems.
This is followed by a brief description of the existing and emerging
standards in the field of digital image and video compression and their
scopes and functionalities. Next, the three main components of compression
schemes, namely, transformation or decomposition, quantization, and symbol
modeling and encoding are presented. The topic of symbol encoding is next
studied in greater detail. In particular, the concept of a Markov source
model and its entropy are presented, and the various coding strategies such
as Huffman coding, arithmetic coding, and LZW coding are studied and their
relative merits and shortcomings are compared. The first day is concluded
with a detailed study of lossless image compression schemes which include bit-
plane encoding, lossless DPCM used in the JPEG lossless standard, and the
Rice algorithm. Finally, a description of the current ITU-T (formerly CCITT)
facsimile standards and the emerging JBIG standard is provided.
The second day contains a detailed description of lossy compression
techniques for still-images which also form the basis for video compression
schemes such as MPEG and H.261. First a brief description of various
quantization strategies such as the uniform scalar quantizer, the Lloyd-Max
nonuniform scalar quantizer, the entropy-constrained scalar quantizer, vector
quantization (VQ), and the emerging technique of trellis-coded quantization
(TCQ) is provided and their relative performances are compared. This is
followed by an analysis of transform image coding schemes with a particular
emphasis on the discrete cosine transform (DCT). The baseline, extended, and
enhanced modes of the JPEG international standard for the compression of
continuous-tone still color images are studied in detail. This is followed by
a study of wavelet and subband coding schemes and their performance merits
compared to DCT. The basic notions of fractal image compression are also
reviewed. Finally, progressive image transmission techniques and various
strategies for constructing image hierarchies are discussed. Topics include
the S-transform, Knowlton's technique, the Laplacian pyramid and the Kodak
Photo-CD image pyramid. This concludes the second day.
The third day is devoted to video compression. Interframe motion information
is a fundamental component of video compression since it facilitates the
utilization of temporal redundancies that naturally exist in video sequences.
Further, it plays an important role in designing efficient pre-compression
algorithms, such as noise filtering. Noise suppression via pre-filtering
greatly increases the efficiency of subsequent compression and is often one
of the differentiating factors used in evaluating a complete compression
system. We develop a unifying framework for fundamentals of motion
estimation and present an overview of motion estimation algorithms within
this framework. Next, we discuss motion-adaptive algorithms for preprocessing
of video. Following a brief overview of different approaches to filtering of
image sequences, we describe two recently developed spatiotemporal filters
for motion-adaptive noise suppression. We next provide a detailed discussion
of both the fundamental and working principles of MPEG-1 and MPEG-2 video
compression standards. We also present a brief overview of H.261 standard
and point out its similarities and differences with MPEG-1. An overview of
available silicon, software, and hardware implementations of the standards
is also furnished. A summary of emerging MPEG-4 activities, aimed at
developing compression standards that are amenable to content-based access
and manipulation of audiovisual information, is followed by an overview of
standardization efforts for digital standard television and high-definition
television (HDTV) in US.
Bulletized Course Outline
Introduction
Product examples
Statistical redundancy and perceptual irrelevancy
Lossless vs. lossy compression
Standards: JPEG, MPEG, H.261, etc.
Compression basics: transformation, quantization, symbol coding
Symbol Encoding
Markov modeling and entropy
Huffman coding
Arithmetic coding
LZW coding
Lossless Compression Techniques
Bit-plane coding
DPCM
Rice algorithm
Facsimile standards (ITU-T Group III and IV, JBIG)
Quantization Strategies
Uniform scalar
Nonuniform scalar (Lloyd-Max)
Entropy-constrained
Vector quantization (VQ)
Trellis-coded quantization (TCQ)
Lossy Compression Schemes
Predictive coding (DPCM)
Discrete cosine Transform (JPEG Baseline, extended, and enhanced modes)
Wavelet and subband coding
Fractals
Image Hierarchies and Progressive Transmission
Variable-amplitude hierarchies (bit-plane encoding, etc.)
Variable-resolution hierarchies (Knowlton's technique, S-transform, etc.)
Photo-CD compression
Introduction to Video Compression
Motivation for video compression
Interframe and Intraframe approaches to video compression
Motion estimation
Fundamental principles of motion estimation
Overview of algorithms (block matching, hierarchical block matching, etc.)
Noise Suppression
Motion-detection based approaches
Motion-compensated approaches
Adaptive, motion-compensated spatiotemporal filters
Scratch detection and removal
Video Compression Standards
The MPEG-1 Standard
Summary of the H.261 Standard
The MPEG-2 Standard
The future: MPEG-4 activities
Examples of silicon, hardware board and software implementations
Overview of Advanced Television
Advanced TV and HDTV
Brief history of HDTV standardization in US
Outline of The Grand Alliance System
Instructors' bios:
Majid Rabbani received his M.S. and Ph.D. degrees in electrical
engineering from the University of Wisconsin in Madison in 1980 and
1983, respectively. He joined the Eastman Kodak Research Laboratories in
1983, where he is currently a research associate and the head of the
image compression Technology Area within the Imaging Science Division.
He is also involved in many educational activities among which are
teaching graduate courses at the RIT (Rochester Institute of Technology)
EE department, satellite courses for NTU (National Technological
University), and short courses for MIT, RIT and various technical
Societies.
Dr. Rabbani is the recipient of the 1988 Kodak C. E. K. Mees
Award and the co-recipient of the 1990 Emmy Engineering Award (for
image compression research) in recognition of the
Still-Video Tranceiver System. He directed the efforts of the Los Alamos
team in charge of the digital enhancement of the Rodney King beating
video tape and subsequently testified as an expert witness for the case
in 1993. His current research interests span the various aspects of
digital signal and image processing where he has published over 40
technical articles and holds 10 patents. From 1990-1994 he was the
Editor of the Journal of Electronic Imaging. He is a Fellow of SPIE, and
a senior member of IEEE. He is the coauthor of the book Digital Image
Compression Techniques published by SPIE Press in 1991 and the editor
of the SPIE Milestone Series on Image Coding and Compression ,
published in 1992.
M. Ibrahim Sezan received the B.S degrees in Electrical Engineering and
Mathematics from Bogazici University, Istanbul, Turkey in 1980, with the
highest honors. He received the M.S degree in Physics from Stevens Institute
of Technology, Hoboken, New Jersey, and the M.S and Ph.D degrees in
Electrical Computer and Systems Engineering from Rensselaer Polytechnic
Institute, Troy, New York in 1982, 1983 and 1984, respectively. Since 1984,
he is with Eastman Kodak Company, Rochester, New York. Presently,
he leads the Motion and Video Technology Area in the Imaging Research and
Advanced Development Laboratories. He also holds an adjunct faculty position
at the Electrical Engineering Department at the University of Rochester.
Dr. Sezan was the co-recipient of the A. B. Du Mont award at Rensselaer
Polytechnic Institute in 1984. During 1988-1992, he served as an Associate
Editor of the IEEE Transactions on Medical Imaging. From 1992 to 1994 he
was an Associate Editor of the IEEE Transactions on Image Processing. He
contributed to the books Image Recovery: Theory and Application
(Academic Press, 1987), Mathematics in Signal Processing (Oxford
1987), Handbook of Signal Processing (Marcell Dekker, 1988),
Digital Image Restoration (Springer Verlag, 1991), Real-Time
Optical Information Processing (Academic Press, 1994) and edited
Selected Papers in Digital Image Restoration (SPIE Milestone Series,
1992). He is the co-editor of the book Motion Analysis and Image
Sequence Processing (Kluwer, 1993). His research interests include
video analysis, processing and compression, image restoration and
enhancement, and digital image and video libraries. Dr. Sezan is an active
participant in the MPEG standards; he actively publishes and teaches in the
area of image and video processing.
Registration Information
Dates: Wednesday, September 13 - Friday, September 15, 1995
Times: Regular sessions will begin at 8:30 AM and end at 5:00 PM.
Refreshments will be served at 8:00 AM daily and at breaks. Lunch
periods are from 12:00 to 1:30.
Location: Classes will be held at Portland State University
(PSU), located in downtown Portland. Detailed information will be
sent to registrants.
Fees:
Early Registration: $895. Registration form and payment
must be postmarked BY August 17.
Late Registration: $995. Registration form and payment
postmarked AFTER August 17.
All registration materials must be received BY August 31. Fees
include lecture, course materials, refreshments and a Certificate of
Completion.
A 15% discount will be granted when 6 or more people from the same
company register for the course.
Refund: A full refund will be given for cancellations received by phone at
PSU, (503) 725-3806, prior to August 31. No refund will be made for
a cancellation notice received after August 31, or for
non-attendance. A substitute may attend in place of the registered
participant.
Accomodations: For reservations made before 8/27/95, special rates for participants are arranged with:
Airlines-Carlson Travel :(800) 624-4865
\ \ 5% Airline Discount (Credit Card Only)
\ \ Ask for Renee
Days Inn City Center :(800) 899-0248
\ \ $59 Single, $64 Double, $69 Double-Double
Red Lion, Portland Center :(503) 221-0450
\ \ $95 Single, $110. Double
Mention Digital Imaging when making reservations. Both hotels are a
pleasant 10 minute walk to PSU. Buses and MAX trains in the downtown
area are free, so it should not be necessary to rent a car. Other
nearby hotels:
The Benson: (503) 228-2000
Heathman Hotel: (503) 241-4100
Hilton Hotel: (503) 226-1611
Mallory Hotel: (503) 223-6311
Marriott Hotel: (503) 226-7600
Excursions: Spousal/Companion activities will be arranged depending on
interest.
WORKSHOP REGISTRATION
Digital Image and Multimedia Compression:
Fundamentals and International Standards
September 13 - 15, 1995
Portland State University, Portland, Oregon
Name:
Company:
Address:
City/State/Zip:
Email address: Work Phone:
Payment Options:
$895 if postmarked BY August 17, 1995
$760 per person for six or more registrations from one company
$995 if postmarked AFTER August 17, 1995
$845 per person for six or more registrations from one company
Enclosed is a check made payable to
Portland State University - E.E. Dept.
Please charge the registration fee to my credit card:
Visa MC
Account # Exp. Date
Signature:
I would like to receive information on the Spousal/Companion activities.
MAIL OR FAX FORM to:
Portland State University
Department of Electrical Engineering
Digital Image and Multimedia Compression
P.O. Box 751
Portland, OR 97207-0751
FAX: (503)725-3807
Phone: (503)725-3806
Email: laura@ee.pdx.edu
All registration materials must be received by August 31, 1995
For more information, please see our WWW homepage at
http://www.ee.pdx.edu/short_courses/image_compression/
------------------------------
End of VISION-LIST digest 14.26
************************
