Copy Link
Add to Bookmark
Report
VISION-LIST Digest Volume 11 Issue 26
VISION-LIST Digest Mon Jul 20 12:03:15 PDT 92 Volume 11 : Issue 26
- Send submissions to Vision-List@ADS.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@ADS.COM
- Access Vision List Archives via anonymous ftp to FTP.ADS.COM
Today's Topics:
Re: CCD sensor for X-rays
Motorized Lenses WITH feedback
Re: Need OCR help
Graduate Research Assistantships
Position available - imaging and image analysis developer
FAQ request / Stereo Vision references
Request for refs on document analysis and visual tracking
CALL FOR PAPERS
Conference announcement
CFP: 3rd Int. Wkshp on Frontiers in Handwriting Recognition
CFP: Asia Pacific Engg Journal
Tech Report
----------------------------------------------------------------------
Date: Thu, 16 Jul 92 11:39:17 CDT
From: schultz@halley.serc.3m.com (John C. Schultz)
Subject: Re: CCD sensor for X-rays
CCD's (and Si) sensitivity do not extend into the X-ray region (near
UV is about it).
However you can readily buy flourscent screens and aim a TV camera at
the screen - check with medical suppliers. Alternatively, a company
called Lixi, Inc 1438 Brook Drive, Downers Grove, ILL, USA 60515
312-620-4646 sells a nice, small X-ray, screen and sensor package.
Unfortunately they cost around $25,000US.
------------------------------
Date: Thu, 16 Jul 1992 15:07:34 GMT
From: srsuresh@laser1.engr.utk.edu (Marapane Suresh)
Organization: University of Tennessee ECE dept.
Subject: Motorized Lenses WITH feedback
Keywords: Motorized Lenses, Feedback
Hello:
I am interested in purchasing a couple of lenses that have
electric motors WITH feedback to operate zoom, focus, and
iris. Does anyone know where I can buy such lenses?
Thanks in advance.
Suresh
[ This is of great interest a many of us, please post responses directly
to the List. phil... ]
|| Suresh B. Marapane | srsuresh@vms1.engr.utk.edu ||
|| Research Associate | srsuresh@laser1.engr.utk.edu ||
|| Computer Vision and Robotics | ||
|| Research Laboratory | Office: (615) 974 - 5493 ||
|| Dept. of Elec. & Computer Engr. | Lab.: 974 - 5476, 4876 ||
|| 409 Ferris Hall | Fax: 974 - 5459 ||
|| Knoxville, TN 37996-2100 | ||
------------------------------
Date: Thu, 16 Jul 1992 14:57:39 +0100
From: arms@cs.ualberta.ca (Bill Armstrong)
Organization: University of Alberta, Edmonton, Canada
Subject: Re: Need OCR help
Vision-List@ads.com writes:
> Optical Character Recognition- Can anyone help?
>
> A colleague here is wanting to do some OCR,
> Can anyone suggest a good lead-in paper to practical techniques?
If you want to try adaptive logic networks, there is a ready-made OCR
demo with C and C++ source code for Windows 3.x available via
anonymous ftp from menaik.cs.ualberta.ca [129.128.4.241] in
pub/atre26.exe. This self-extracts to a new subdirectory of the main
directory. If you want icons, run setup in the latter directory.
ALNs provide great immunity to noise and distortion. You can see that
by hand-drawing in some test characters. The system can learn to
recognize any characters you give it. Once you've used the Windows
version and used its on-line help, you could use the Unix version for
developing a system for another machine (Sparc, Amiga,...)
Prof. William W. Armstrong, Computing Science Dept.
University of Alberta; Edmonton, Alberta, Canada T6G 2H1
arms@cs.ualberta.ca Tel(403)492 2374 FAX 492 1071
------------------------------
Date: Thu, 16 Jul 1992 14:57:58 GMT
From: srsuresh@laser1.engr.utk.edu (Marapane Suresh)
Organization: University of Tennessee ECE dept.
Subject: Graduate Research Assistantships
Keywords: Assistantships, Vision, Robotics, Graphics
GRADUATE RESEARCH ASSISTANTSHIPS
The Computer Vision and Robotics Research (CVRR) Laboratory of
the University of Tennessee, Knoxville is soliciting applications from
qualified students for Ph.D. studies in the areas of:
1. Sensor-Based Intelligent Robotics, emphasizing integrated systems.
2. Human-Machine Systems, including interactive graphical interfaces
and virtual reality based control.
3. Image Understanding, emphasizing quantitative preattentive vision,
computational frameworks for object recognition and active perception.
Our research currently involves several research faculty members and
about two dozen graduate students. The primary academic program is
situated in the Electrical and Computer Engineering Department, however
faculty members from Mechanical Engineering, Psychology
(Perceptual and Human-Factors), and Oak Ridge National Laboratory
provide research guidance, instruction, and supervision. Sponsors of our
work include major Government Programs (DOE, DOD, Army, NSF)and
Industry (U.S and Japan).
The CVRR Lab maintains a modern, well equipped facility with several
unique capabilities. The computational resources include a minicomputer,
parallel computer, and several graphics workstations, real-time image
acquisition and processing hardware, high speed peripherals, an
industrial robot endowed with a variety of external sensors, a mobile
manipulator system. The sensory devices incorporated include vision,
ultrasonic range, laser point range, 3-D laser imaging system, thermal
imaging system, touch, force-torque, proximity, sound, and voice
recognition system.
Research assistantship package includes competitive stipend and full
tuition waivers. Prospective students must be highly motivated and have
exceptional academic qualifications. Prior research experience is
desirable. Applicants interested should submit resume, listing major
courses, GPA and GRE scores and references, to srsuresh@laser1.engr.utk.edu
or:
Mohan M. Trivedi, Professor
Computer Vision and Robotics Research Laboratory
Electrical & Computer Engineering Department
The University of Tennessee
Knoxville, TN 37996-2100
e-mail: trivedi@vms1.engr.utk.edu
------------------------------
Date: Fri, 17 Jul 92 08:08:27 CDT
From: zaxw16@nap.amoco.com (Amir Waks)
Subject: Position available - imaging and image analysis developer
A challenging openning is currently available in my group at Amoco Technology
Company. We are looking for a professional who has good understanding of
image analysis techniques as applied for bio-medical applications, and
particularly, has also very good software engineering and design skills.
Position Available: Scientific Developer - Imaging and Image Analysis
At the Biotechnology Department in Amoco Research Center we design and
develop full scope imaging systems for medical diagnostics. The development
efforts include the following domains: sensitive image acquisition, optics and
precise mechanics, reliable image analysis, and software engineering.
The Biotechnology Department has an excellent career position for a scientific
programmer in the areas of imaging and image analysis. The qualified
individual will be responsible for devising and implementing reliable
algorithms for object segmentation and recognition, testing the algorithms on
a large set of image data, modifying the algorithms accordingly, and
evaluating the system performance. An ability to understand the trade-off
between the various domains of the imaging system, and to identify the
optimal design is needed in this position. Strong software development skills
are essential since the qualified candidates will be responsible for setting
and writing a stand alone application.
The successful candidate should possess an BS/MS in Computer Science,
Electrical Engineering, Computer Engineering, or related fields. Additionally,
the candidate should possess two years experience in applying statistical
and/or morphological image analysis, and strong mathematical skills are
advantageous. Proficiency in segmentation techniques is needed and
knowledge of Pattern Recognition is desired. A minimum of two years
experience with developing software in C or C++ for the Macintosh with
MPW or Think C is needed. Candidates with at least two years experience
using Sun systems with Open Look and Motif will be considered too.
Successful applicants must be able to cooperate in a team environment and
to accomplish their tasks independently.
Amoco offers a competitive salary, a comprehensive benefits package plus a
professional and congenial work environment. For immediate consideration,
please forward your resume and salary history in confidence to the
undersigned.
Amir Waks, Ph.D
Amoco Technology Company
Mail Code F-2
P.O. Box 3011
Naperville, IL 60566
------------------------------
Date: Thu, 16 Jul 1992 08:26:46 GMT
From: oscar@mi.el.utwente.nl (Oscar ter Hofstede - TUT)
Organization: Utwente
Subject: FAQ request / Stereo Vision references
I am looking for a book (preferably) or article that gives a recent
overview on Stereo Vision, ideas and algorithms.
I have found a reference to a recent book:
Ayache, N.
Artificial vision for mobile robots: stereo vision and
multisensory perception.
MIT Press, 1991.
Does anyone know this book who can tell me if it is worth purchasing?
Furthermore, I would like to know where I can find a FAQ for this
group (assuming there is one).
Thanx,
Oscar.
Oscar ter Hofstede Tel: +31-53-892778 Fax: +31-53-340045
Faculty of electrical engineering -- University of Twente, The Netherlands
------------------------------
Date: Fri, 17 Jul 1992 18:15:26 UTC
From: lbaumela@mayor.dia.fi.upm.es
Subject: Request for refs on document analysis and visual tracking
Dear coleagues,
I am trying to make a list of good papers, books, etc.. on two subjects
of my interest:
- DOCUMENT ANALYSIS AND RECOGNITION.
- VISUAL TRACKING
interesting references on any of these subjects are highly appreciated.
Comments on the reference you send are also wellcome.
Thanks in advance.
PD: If anyone is interested on any of these lists, let me know and I will
will forward the information I get.
Luis Baumela.
Facultad de Informatica | Computer Science School
Universidad Politecnica de Madrid (UPM) | Madrid Tech. University.
Campus de Montegancedo
28660 Boadilla del Monte
Madrid (SPAIN)
Tel +34 1 336 74 39
Fax +34 1 352 48 19
e-mail: lbaumela@mayor.dia.fi.upm.es
------------------------------
Date: Friday, 17 July 1992 09:35:47 EDT
From: Dave.McKeown@MAPS.CS.CMU.EDU
Subject: CALL FOR PAPERS
PLEASE NOTE:
DO NOT Reply To The Mailing List cv@cs.uiuc.edu
Send Questions to dmm@cs.cmu.edu
SPIE OE/Aerospace Sensing
Orlando '93
Orlando Marriot World Center Hotel
Orlando, Florida
12-16 April, 1993
C A L L F O R P A P E R S
SPIE 93 Conference:
Integrating Photogrammetric Techniques With
Scene Analysis and Machine Vision
April 14-15 1993
Conference Chairs
Eamon Barrett Lockheed Missiles and Space Company
David McKeown Carnegie Mellon University
Conference Co-Chairs
Michael Brill Science Applications International Corp
Franz Leberl Vexcel
George Lukes U.S. Army Topographic Engineering Center
J. Chris McGlone Carnegie Mellon University
Edward Mikhail Purdue University
Toni Schenk Ohio State University
The focus of this conference is the application of rigorous photogrammetric
techniques to automated scene analysis and image understanding with primary
emphasis on the analysis of remotely sensed imagery. Many automated and
semi-automated image interpretation algorithms have reached a level of
refinement where the introduction of accurate photogrammetry may be essential
to successful results. The addition of photogrammetric knowledge to these
algorithms will also provide additional information and constraints. We
solicit contributions to this conference that focus on aspects of
photogrammetry and computer-assisted image interpretation in the following
areas:
(1) Use of photogrammetry in computer vision algorithms for object
recognition, modeling, and scene analysis including:
o Use of photogrammetric information to guide computer vision algorithms.
o Use of high quality geometric information and invariants.
o Precision of derived points and features.
o Managing uncertainty in derived models.
o Computer vision in typical photogrammetric problems:
resection
calibration
tie-point determination
(2) Automated mapping applications including:
o Terrain modeling.
o Automated image-scene registration.
o Cartographic feature extraction of manmade structures.
o Understanding and utilization of existing maps, databases, imagery.
o Surface cover description and characterization.
o Cartographic change detection.
o Applications to forestry, land use, environmental monitoring.
(3) Multisensor fusion using explicit photogrammetric models including
modeling of non-traditional sensors, such as:
o Airborne multi-spectral scanners
o Imaging radar
o FLIR, IR
o Sonar
(4) Design issues for advanced photogrammetric workstations including:
o Effects of image handling, resampling, and softcopy stereo display on
measurment precision.
o Automation vs. operator interaction.
o Output verification and editing.
(5) GIS issues in digital photogrammetry including:
o Utilization and update of databases.
o Types of derived products.
o Maintenance of spatial data over time.
Paper submissions should be approximately 15 pages in length and should clearly
address one or more of the topics above. Extended Abstracts can be submitted
for review. These should be between 500 - 1000 words in length and should
clearly convey the approach and results of the research. Papers will be
selected for either presentation during a regular session, or in a poster
presentation. Please indicate a preference for presentation.
Please send (3) copies of full papers or extended abstracts to:
David M. McKeown
Digital Mapping Laboratory
School of Computer Science
5000 Forbes Avenue
Carnegie Mellon University
Pittsburgh, PA. 15213
USA
dmm@cs.cmu.edu
(412) 268-2626
(412) 681-5739 (FAX)
Deadline for receipt of papers and extended abstracts is September 14, 1992.
Acceptances will be announced by November 15, 1992.
Completed Manuscripts are due by March 15, 1992.
------------------------------
Date: Fri, 17 Jul 1992 09:11:03 +0200
From: Claude.Labit@irisa.fr (Claude Labit)
Subject: Conference announcement
*********************** IMAGE'COM_93 ************************
Communicating with images : from tools to applications
2nd INTERNATIONAL CONFERENCE
DEDICATED TO
IMAGE COMMUNICATION
Bordeaux - France - Palais des Congres
MARCH 23-24-25, 1993
*********************
* *
* CALL FOR PAPERS *
* *
*********************
==================
CONFERENCE OUTLINE
==================
IMAGE'COM_93 is a forum for professionals involved in research,
development, marketing, manufacturing and usage of image
technologies to meet, discuss and understand their objectives,
requirements and constraints.
IMAGE'COM_93 will allow them to establish links between :
- the convergence of the techniques of computers,
telecommunications and audio-visual equipment,
- the proliferation of specific applications and possible
synergy moving towards generic services,
- the definition of socio-economic needs,
- the expectations of application designers and creators.
IMAGE'COM_93 will be more specifically devoted to user interfaces,
and to applications and creation of services.
IMAGE'COM_93 will concern specialists working in the field of
image technologies as well as those more interested in
applications related to these techniques such as medical imaging,
videoconferencing, satellite imaging, television.
As a guideline, papers on the following subjects would be
appreciated :
1 - PRODUCTION, CREATION
- Acquisition, Pre-processing
- Signal sensors and their intelligence, active vision
- Multisensor fusion
- Picture taking equipment (TV cameras, photo...)
- Equipment and software for image synthesis
2 - TRANSFORMATION, HANDLING
- Analysis, coding, compressing
- Specialised processors
- Programmable parallel machines
3 - COMMUNICATION, STORAGE
- Digital image and networks
- Access
- Evolution of image transmission and distribution
- Normative aspects
- Archiving equipment (video recorders, servers...)
4 - MULTIMEDIA
- Virtual Reality
- Private and communicating multimedia
- Normative aspects,
- Growing complexity of terminals (videophones, multifunction
monitors, dedicated personal computers...)
5 - INTERPRETATION
- Image recognition and classification
- Computer vision
- Analysis/synthesis cooperation
- Cognitive aspects, neuronal systems, knowledge models
- Physiology and subjectivity of perception
6 - APPLICATIONS, SERVICES
- Broadcast graphics
- Image synthesis (audiovisual creation, modeling, environment
simulation...)
- Operational systems
- Conception methodology
- Publishing (CAD/CAM, Interactive CD...), artistic
creations
- Medical imaging, videoconferencing, satellite imaging,
television...
===============
CALL FOR PAPERS
===============
Those who wish to contribute to one of these themes should send
their typed summary in English (French translation optional) of
between 600 to 1000 words and one or two illustrations if desired,
to the conference secretariat by 31 July 1992.
===========
PROCEEDINGS
===========
Authors will be informed by the end of October 1992 of the Program
Committee's selection from papers submitted.
For the selected contributions a full length camera ready copy in
English will be required by 31 January 1993.
Authors must produce a text in English accompanied by a summary in
French if so desired.
The selected papers will be published in the conference
proceedings at the organizers discretion.
========
LANGUAGE
========
English or French, but English is preferable. Simultaneous
translation (English/French or French/English) will be provided
during the conference. Speakers are asked to present their audio
and video material (slides, transparencies, ...) in English.
=========
DEADLINES
=========
- 31 July 1992 : deadline for summaries.
- 31 October 1992 : Acceptance of papers.
- 31 January 1993 : deadline for papers.
==========
EXHIBITION
==========
An international professional exhibition on image technologies in
the communciations sector will be held in Bordeaux from 23-24-25
March 1993.
===========
SECRETARIAT
===========
Authors should give their complete address (including fax) and
send their summary to :
ADERA
IMAGE'COM_93
BP48
33166 Saint-Medard-en-Jalles CEDEX
FRANCE
Phone : (+33) 56 70 68 56
Fax : (+33) 56 95 98 83
==========================
RESULTS OF THE 1st EDITION
==========================
IMAGE'COM_90 took place on 19-20-21 November 1990 in Bordeaux.
This first edition brought together over 280 delegates including
over 50% coming from 16 overseas countries.
During the conference 80 papers were presented, including 40 from
16 overseas countries, as well as 15 posters.
=======================
INTERNATIONAL COMMITTEE
=======================
Chairman : D.NAGEL - APPLE COMPUTER (USA)
Secretary : G.FERRIEU - SEE (France)
Past Chairman : P.CONRUYT - FRANCE TELECOM (France)
Constitution pending.
====================
SCIENTIFIC COMMITTEE
====================
Chairman : J.GENINET - Philips EGP (France)
Secretary : E.PETITHOMME - FRANCE TELECOM (France)
Members : H.AEBISCHER - APPLE COMPUTER EUROPE (France)
K.ANJYO - Hitachi Ltd (Japan)
B.BOUCHARE - CNET Lannion (France)
R.BOYER - THOMSON LER (France)
G.BRUN - TDF (France)
G.CAMBILLAU - SNCF (France)
V.CANTONI - Univ. di Pavia (Italy)
A.FOURNIER - Univ. of British Columbia (Canada)
H.FUCHS - Univ.of North Carolina (USA)
A.GAGALOWICZ - INRIA (France)
B.E.HASKELL - AT&T Bell Labs (USA)
F.HOTTIER - PHILIPS LEP (France)
M.IMBERT - Universite de Paris VI
J-B.JACOB - ALCATEL CIT (France)
J-C.JOLIVET - SAT (France)
C.LABIT - IRISA (France)
H.LANGE - Philips (Germany)
M-J.LEFEVRE - CNES (France)
H.MAITRE - SUP TELECOM PARIS / ENST (France)
S.K.MUN - Georgetown Hosp. Univ. (USA)
J.E.NAVARRO - Alcatel Standard Electrica (Spain)
S.OKUBO - NTT / Human Interface Labs (Japan)
A.POIGNET - CCETT (France)
R.POUGHEON - BULL France)
H.PRECIOSO - IBM (France)
P.PURCELL - University of Ulster (UK)
L-M.REIBER - Thomson C.E. (France)
J.SANZ - IBM/Almaden Research Center (USA)
J.SHEN - Univ. Bordeaux III (France)
J.SKRZYPEK - UCLA (USA)
G.STAMON - EHEI (France)
S.TANIMOTO - Univ. of Washington (USA)
P.J.W.TEN HAGEN - CWI (Netherlands)
D.THALMANN - EPFL (Switzerland)
M.TRIBOULET - Raynet International (Belgium)
C.VEYRES - FRANCE TELCECOM / STI (France)
B.VOETEN - Alcatel Bell Telephone (Belgium)
W.J.WELSH - BT Laboratories (UK))
E.WILSON - UER (Geneva)
B.ZAVIDOVIQUE - ETCA (France)
==========
ORGANISERS
==========
IMAGE'COM_93 is organised by :
- SEE : Societe des Electriciens et Electroniciens
- IREST : Institut de Recherches Economiques et Sociales sur
les Telecommunications
- ADERA : Association pour le Developpement de l'Enseignement
et des Recherches aupres des Universites, des Centres
de Recherche et des Entreprises d'Aquitaine
====================
ORGANISING COMMITTEE
====================
Chairman : J.POUFET - FRANCE TELECOM (France)
Secretary : J.RIVENC - ADERA (France)
Constitution pending.
------------------------------
Date: Mon, 20 Jul 92 12:04:50 EDT
From: hull@cs.Buffalo.EDU (Jon Hull)
Subject: CFP: 3rd Int. Wkshp on Frontiers in Handwriting Recognition
C A L L F O R P A P E R S
Third International Workshop on
Frontiers in Handwriting Recognition
May 25-27, 1993
Buffalo, New York, USA
Recognition of handwriting by computer is a rapidly
expanding area of pattern recognition and artificial intel-
ligence research. The third international workshop, devoted
to advances in this topic, will be hosted by CEDAR, the
United States Postal Service Center of Excellence for Docu-
ment Analysis and Recognition located at the State Univer-
sity of New York at Buffalo. Previous workshops in this
series were held in Montreal, Canada in 1990, and in Bonas,
France in 1991.
The workshop will address the goal of designing and
building systems that recognize and understand handwritten
characters, words, symbols, and text. We hope to bring
together an interdisciplinary group of researchers who can
discuss issues of online (pen-based) and off-line (digitized
images of previously recorded script) handwriting recogni-
tion. We are especially interested in the participation of
those researchers who have developed systems for specific
domains or have innovative applications to report. Facili-
ties for research demonstrations and a vendor exhibit will
also be provided.
Topics of Interest Include:
% Recognition of characters, digits, words, text, and
symbols (both on-line and off-line)
% Preparation and treatment of training data and testing data
% Feature extraction and optimal selection
% Classification techniques and decision systems
% Segmentation of lines, words, characters, and digits
% Contextual processing and string matching
% Interactive systems
% Multiple recognition algorithms and techniques
% Intelligent systems with learning capabilities and
self-correction
% Case studies and innovative applications
% Layout analysis of handwritten and mixed documents
% Use of linguistic information
% Architectures
Submission Requirements
Please mail your submission to arrive by November 30, 1992.
Acceptances will be mailed by December 31, 1992.
Final camera-ready manuscripts will be required before February 28, 1993.
Prospective authors should submit four copies of the following material:
a. 5-10 page summary (8.5" 4 11" single-spaced) outlining
important results and promising work in progress,
problems encountered, and challenges ahead
b. full address, electronic mail, telephone and facsimile
numbers, of the author to whom correspondence must be
addressed
c. description of the data used in your research
Requests for information about registration and
attendance, as well as paper submissions should be sent to:
Rita Hamilton
CEDAR, 226 Bell Hall
Buffalo, NY 14260-0001 USA
e-mail: hamilton@cedar.buffalo.edu
Electronic and facsimile paper submissions cannot be accepted.
ORGANIZING COMMITEE
General Chairman:
Sargur (Hari) Srihari
CEDAR, USA
Local Arrangements/Progam
Richard Fenrich Venu Govindaraju Jonathan Hull
CEDAR, USA CEDAR, USA CEDAR, USA
Program Committee:
Mindy Bokser Radmilo Bozinovic Richard Casey
Calera Recognition Apple Computer, USA IBM Almaden Research
Systems, USA Center, USA
Andy Downton Sebastiano Impedovo Guy Lorette
University of Universita de Bari, IRISA, France
Essex, UK Italy
Shunji Mori Rejean Plamondon Jean-Claude Simon
Ricoh Corporation, Ecole Polytech. Univ. Pierre et
Japan Montreal, Canada Marie Curie, France
Ching Suen Charles Tappert Hans-Leo Teulings
Concordia University, IBM-TJ Watson Res. Nijmegen Institute,
Canada Center, USA Netherlands
------------------------------
Date: Sat, 18 Jul 92 9:54:49 SST
From: elesrini@nuscc.nus.sg (SRINIVASAN (A/P))
Subject: CFP: Asia Pacific Engg Journal
ASIA PACIFIC ENGINEERING JOURNAL Part A
The ASIA PACIFIC ENGINEERING JOURNAL is an international journal
published by the National University of Singapore and distributed by
World Scientific Publishing Co.. Part A is devoted to Electrical
Engineering. The first issue was published in September 1991. There are
FOUR issues every year each covering one of the following areas:
Microelectronics
Communications Engineering
Computer Engineering
Control and Automation
An international editorial board has been constituted for each area.
The members of the board for Computer Engineering include
R Chin (U Wisconsin)
B V Kumar (CMU)
B Liu (Princeton)
E S Kuh (UC Berkeley)
S Tsuji (Osaka).
The next issue in Computer Engineering is scheduled for publication in
June 1993. Tentatively, we would like this to be a Special Issue on
vision and image processing (including neural network techniques
)
We welcome contributions to this special issue. The paper deadline is
December 31, 1992.
You may obtain further details or indicate your interest in submitting a
paper by e-mail to the issue editors ASAP.
Dr V Srinivasan ( elesrini@nusvm.bitnet OR elesrini@nuscc.nus.sg)
Dr SH Ong ( eleongsh@nusvm.bitnet )
Department of Electrical Engineering
National University of Singapore
Singapore
------------------------------
Date: Thu, 16 Jul 92 12:00:08 EDT
From: shah@eola.cs.ucf.edu (Mubarak Shah)
Subject: Tech Report
The following tech report is available for anonymous ftp.
Please post this in vision list, thanks.
MRF-Based Motion Estimation and Segmentation
Yu Tian and Mubarak Shah
CS-TR-92-13
Computer Science Department
University of Central Florida
Orlando, FL 32816
Abstract
In the general structure-from-motion (SFM) problem involving several
moving objects in a scene, the essential first step
is to segment different moving objects.
The conventional gradient-based motion estimation approaches
impose a global smoothness constraint on the image, which results in inaccurate
optical flow estimates at the vicinity of motion boundaries.
In this paper we attempt to deal with, in parallel, the problem of cal flow
estimation and motion segmentation.
We introduce line processes and a piecewise smoothness constraint
to get a more accurate motion estimation.
A more general approach based on the Markov Random Field Models and the mean
field techniques are presented.
It is interesting to note that Horn and Shunck's method \cite{Horn81} becomes a special case of our scheme when the line processes are set to zero.
To deal with the problem of larger motion, we propose an adaptive multigrid approach.
Experiments based on real and synthetic data indicate that our approach provides good estimates
of optical flow and motion boundaries.
%ftp osceola.cs.ucf.edu
Connected to osceola.cs.ucf.edu.
220 osceola.cs.ucf.edu FTP server (SunOS 4.1) ready.
Name (osceola:shah): anonymous
331 Guest login ok, send ident as password.
Password:
230 Guest login ok, access restrictions apply.
ftp> cd incoming
250 CWD command successful.
ftp> ls
200 PORT command successful.
150 ASCII data connection for /bin/ls (132.170.108.100,1083) (0 bytes).
book1.1.ps.Z
book1.2.ps.Z
book1.3.ps.Z
book1.4.ps.Z
clock.ps.Z
clock.tar.Z
paper1.ps.Z
226 ASCII Transfer complete.
94 bytes received in 1 seconds (0.089 Kbytes/s)
ftp> get paper1.ps.Z
200 PORT command successful.
150 ASCII data connection for paper1.ps.Z (132.170.108.100,1084) (463277 bytes).
226 ASCII Transfer complete.
local: paper1.ps.Z remote: paper1.ps.Z
464998 bytes received in 13 seconds (36 Kbytes/s)
ftp> quit
221 Goodbye.
% uncompress paper1.ps.Z
% lpr -Plw3 paper1.ps
------------------------------
End of VISION-LIST digest 11.26
************************
