Date: 25 Jun 93 10:47:44-PST
From: Vision-List moderator Phil Kahn <Vision-List-Request@TELEOS.COM>
Errors-to: Vision-List-Errors@TELEOS.COM
Reply-to: Vision-List@TELEOS.COM
Subject: VISION-LIST digest 12.28
To: Vision-List@TELEOS.COM

VISION-LIST Digest    Fri Jun 25 10:47:44 PDT 93     Volume 12 : Issue 28

 - ***** The Vision List has changed hosts to 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 FTP.TELEOS.COM

Today's Topics:

 Stereo Database
 Optical flow programs and data
 Geometric Hashing
 Frame-by-frame image capturing
 Multilayer HNN
 Research posts in functional/parallel/vision project
 TR: Image Analysis Using Parametric Models
 A short course on Fuzzy Logic
 Workshop: Shape Recognition in 2-D and 3-D Images

----------------------------------------------------------------------

Date: Thu 24 Jun 93 16:34:14-PDT
From: Bob BOLLES <BOLLES@ai.sri.com>
Subject: Stereo Database

People interested in stereo data ...

	We recently transfered the JISCT database of 44 image pairs
(and a few example result files) to the Vision List Archive so that
anyone who's interested can access them.  We collected these images
from JPL, INRIA, SRI, CMU, Teleos and the VAP Program sponsored by
CNES in France.  In any published use of these images, please include
reference to this database and its contributors.

	Included below is an example of how to access these images and
their associated documentation files.

	If you have any problems logging onto the Teleos System or
accessing the Vision List Archive, please contact Phil Kahn at Teleos
(Vision-List-Request@TELEOS.COM).  

	If you have trouble reading the image files, please contact
Harlyn Baker (baker@ai.sri.com) or Bob Bolles (bolles@ai.sri.com).

	We are interested in the number of people using the data, so
please send an E-mail message to Harlyn Baker (baker@ai.sri.com) when
you copy the database, stating your plans for using it.

	If you have any suggestions for future stereo data
acquisitions or have data to contribute, please contact Bob Bolles
(bolles@ai.sri.com).  Hopefully in the future we will be able to
supply more information about camera calibrations and ground truth.

					... Harlyn Baker & Bob Bolles


Typical interaction sequence for getting the data and documentation files:


>ftp ftp.teleos.com
  Name (ftp.teleos.com:baker): anonymous
  Password: <use your username@hostname>
  230 Guest login ok, access restrictions apply.
ftp> cd VISION-LIST-ARCHIVE/IMAGERY/JISCT
  250 CWD command successful.
ftp> ls
  README
  images
  jisct-reader.c
  results
ftp> get README		(copy the documentation file)
ftp> get jisct-reader.c	(copy the C file reader)
ftp> cd images		(access the images subdirectory)
ftp> binary		(switches to binary transfer mode ... vs. ASCII)
ftp> prompt		(toggles (turns off) verification request mode)
ftp> mget *.*.Z		(get the compressed image files)
ftp> cd ../results	(access the results subdirectory)
ftp> mget *.* 		(get the uncompressed result files)
ftp> quit

One then would 
>uncompress *.*.Z	(to uncompress the image files)

------------------------------

Date: Thu, 24 Jun 93 14:14:47 EDT
From: John Barron <barron@csd.uwo.ca>
Subject: Optical flow programs and data

David Fleet, Steven Beauchemin and myself have recently completed the
implementation of a number of optical flow algorithms (see CVPR 92 and
IJCV (in press)). These programs and our data were posted to
vision-list some time ago but since then we have made some changes to
3 of the programs (lucas.c horn.c and nagel.c) and some of the image
sequences. Hence, we often obtain better results than those reported
in CVPR or in a TR. If you are using any of these programs please make
sure you have the most up-to-date versions. A revised version of the
TR will be appearing soon. Anyone who has received a copy of the old
TR is automatically on the mailing list, otherwise please send me
email with your postal address if you wish a copy.  The programs/data
are available as a tarfile consisting of compressed files. The
programs/data (as individual files) are also available from
ftp.csd.uwo.ca (cd /pub/vision).

------------------------------

Date: Wed, 23 Jun 1993 17:55:49 GMT
From: dutta@scorgies.NoSubdomain.NoDomain (Paul Dutta-Choudhury)
Organization: Cognex Corp.
Subject: Geometric Hashing

One of the more interesting model based matching algorithms to arise
in the past few years is the Geometric Hashing method.  I've collected
a few of the relevant articles on this topic and am wondering if people
have actually implemented prototype systems using the technique.  I
would be interested in hearing different viewpoints on the strengths,
and limitations of the technique.  Specifically I am interested in the
2-D case with scale and rotation invariance, 3-D performance is
interesting but really isn't needed for the majority of current
industrial pattern location applications.

Paul D-C
Cognex Corp.

------------------------------

Date: Tue, 22 Jun 93 21:27:04 EDT
From: jdavis@eola.cs.ucf.edu (James Davis )
Subject: Frame-by-frame image capturing

I'm currently doing research in hand-gesture recognition and
need frame-by-frame digitized images as input to a computer 
program.  

At this time we, our setup is
 
camera/VCR -> SPARCstation w/ vfctool -> image => program on SPARC

This setup requires the user to manually digitize the frames
from the camera/VCR using vfctool.  I'm looking for a unit that 
will do this automatically, such as

camera/VCR -> unit -> frame-by-frame images => program on SPARC

Any information or comments is welcome.

Jim Davis
jdavis@eola.cs.ucf.edu

------------------------------

Date: Thu, 24 Jun 1993 16:03:21 GMT
From: uwechueo@cse.fau.edu (Okechukwu A. Uwechue)
Organization: Florida Atlantic University, Boca Raton, FL, USA
Subject: multilayer HNN

Can anyone direct me to where I can get some info on **multilayer**
higher-order neural nets ? I am researching HNNs for pattern 
recognition,

thanks,
uwechueo@sol.cse.fau.edu

------------------------------

Date: Tue, 22 Jun 1993 13:17:54 GMT
From: greg@cee.heriot-watt.ac.uk (Greg Michaelson)
Organization: Dept of Computing & Electrical Engineering, Heriot-Watt 
              University, Scotland
Subject: Research posts in functional/parallel/vision project

                              Heriot-Watt University

                Department of Computing and Electrical Engineering

              Research Associates in Parallel Processing and Vision

        Applications   are   invited   for   two   SERC-funded    Research
        Associateships  to  work  on  aspects  of  rapid  prototyping  and
        implementation of algorithms for image interpretation on  multiple
        instruction  multiple data (MIMD) architectures.  Although working
        closely  together,  each  RA  will  assume  responsibility  for  a
        specific  programme.   The first will concentrate primarily on the
        software  methodology,  including  functional  specification   and
        instrumentation   of   algorithms   in   Standard   ML  and  their
        transformation into a parallel imperative language, OCCAM 2.   The
        other   will   undertake   the   development,   optimisation   and
        implementation of algorithms for a cooperative  multi-level,  dual
        source  (depth, intensity) parallel system.  The persons appointed
        will join a lively research group working on  several  aspects  of
        computer vision and software development.

        Applicants should  have  an  honours  or  postgraduate  degree  in
        Computer  Science  or a related discipline, together with relevant
        postgraduate experience.  The posts are tenable for  three  years,
        commencing  as  soon  as  possible  after  the  1st  October.  The
        starting salary will be in the range #UK 12,828(pt 3) to #UK 15,796
        (pt 7) depending on age and experience. Enquiries should be directed
        initially to the Staff Officer, Heriot-Watt University, Riccarton,
        Edinburgh  EH14 4AS, from whom further information and application
        forms may be obtained.  The closing date for applications is  31st
        August 1993.

        Informal enquiries may be  directed  to  Andrew  Wallace  or  Greg
        Michaelson    (tel.    031-451-3423/2,    email    -    andy@   or
        greg@cee.hw.ac.uk.)

------------------------------

Date:  Wed, 23 Jun 1993 12:36:48 +0200
From: " (Franc Solina)" <franc@ninurta.fer.uni-lj.si>
Subject:  TR: Image Analysis Using Parametric Models
 
 Recently, Ales Leonardis defended his Ph.D. dissertation in computer vision 
 at University of Ljubljana, Slovenia.
 Advisers were Franc Solina (Univ. of Ljubljana) and
               Ruzena Bajcsy (Univ. of Pennsylvania).
 
 
                        Ales Leonardis
  
               IMAGE ANALYSIS USING PARAMETRIC MODELS:
             Model-Recovery and Model-Selection Paradigm
              (Ph.D. dissertation in computer science) 
 
                          ABSTRACT
 
  The significance of detecting geometric parametric structures has long been
  realized in the vision community. In this thesis, we present a reliable and 
  efficient method for extracting simple parametric geometric structures in 
  images, i.e., parametric surfaces in range images and parametric curves in 
  edge images. Unlike the approaches which attempt to solve the problems 
  in a single processing step using sophisticated means, we propose a method
  which is an effective combination of simple component algorithms.
  
  The method consists of two intertwined procedures, namely model-recovery
  and model-selection. The first procedure systematically recovers parametric
  models in an image creating a redundant set of possible descriptions,
  while the model-selection procedure searches among them to produce an
  optimal result in terms of the objective function.
  
  The reliability of the recovery procedure which builds the parametric
  models is ensured by an iterative procedure through simultaneous
  performance of data classification and parameter estimation.
  The overall relative insensitivity to noise and minor changes in input
  data is achieved by considering many competitive solutions and selecting 
  those that produce the simplest description.
  The selection procedure is defined as a quadratic Boolean problem, and
  the solution is sought by the WTA (winner-takes-all) technique, which 
  turns out to be a good compromise between the speed and the accuracy 
  of the solution.
  
  The presented method is efficient for two reasons: firstly, it
  is designed as a search which utilizes intermediate results as a
  guidance toward the final result, and secondly, it combines model
  recovery and model selection in a computationally efficient procedure.
  
  We have designed and implemented the proposed method which proved to be 
  successful for recovering parametric surface models in range images and
  parametric curve models in edge images.
  
  Key words: Bivariate polynomials, combinatorial optimization, 
  computer vision, curve fitting, edge images, edge image segmentation,
  parametric models, quadratic Boolean problem, range images, range image 
  segmentation, regression updating, scene analysis, surface fitting, 
  surface modeling.
  
  
 For more information and how to ftp a postscript file of the dissertation
 please contact Ales Leonardis:
 
 Ales.Leonardis@fer.uni-lj.si      or        ales@grip.cis.upenn.edu

------------------------------

Date: Wed, 23 Jun 1993 15:48:06 GMT
From: iiscorp@netcom.com (IIS Corp)
Organization: NETCOM On-line Communication Services (408 241-9760 guest)
Subject: A short course on Fuzzy Logic

              Intelligent Inference Systems Corp. 

                         Presents

         Zadeh, Ruspini, Bezdek, Bonissone, and Berenji

                            On

              
                Fuzzy Logic Inference Systems

     
                   A Five Day Short Course
                     San Francisco, CA
                     July 26-30, 1993

Introduced by Lotfi Zadeh, Fuzzy Logic methods may be used to 
design intelligent systems utilizing knowledge expressed in natural 
language.  This methodology, an important source of artificial 
intelligence applications, permits the processing of both symbolic 
and numerical information, being particularly suited to the study of 
systems known only in a vague, imprecise, and uncertain manner.  
Fuzzy logic has been applied to control trains (Sendai subway), 
elevators, household appliances, cameras, and manufacturing 
processes.  Systems designed and developed utilizing fuzzy-logic 
methods have been shown to be more efficient than those based on 
conventional approaches.  In combination with Computational Neural 
Networks techniques, fuzzy-logic methods may be used to design 
robust adaptive control systems.  This course discusses the  
application of fuzzy logic and neural networks techniques to the 
design of fuzzy and hybrid neuro-fuzzy systems.  More than 14 case 
studies will be discussed in detail.  At the completion of this 
course, you will have a full understanding of the benefits of this 
technology, will know  about existing successful applications, and 
will develop the necessary understanding and knowledge to design 
and apply fuzzy logic to your particular needs.

Who Should Attend: 

Engineers, technical managers and project leaders, scientists, 
systems analysts, as well as others who would like to have more 
knowledge about this emerging technology.

About the course: 

This is the strongest and most complete course 
available on this topic. All the presenters of this course are 
pioneers of the field, including Professor Lotfi Zadeh, who 
introduced its seminal ideas and concepts.  Even if you have already 
taken a course or  tutorial on fuzzy logic, you should try to attend 
this course since it will provide you with a deeper understanding of 
the latest techniques and applications in the areas of fuzzy logic, 
soft computing, pattern recognition, intelligent control, 
computational neural networks , and adaptive neuro-fuzzy systems.

Spend 5 days with the pioneers of Fuzzy Logic.

                      Course Outline

Day 1. Fundamentals of Fuzzy logic -- Enrique Ruspini
Day 2: Advanced  Fuzzy Logic and Intelligent Control -- 
	Lotfi Zadeh, Hamid Berenji
Day 3: Numerical Pattern Recognition  -- Jim Bezdek 
Day 4: Approximate Reasoning in Knowledge-based systems 
	-- Piero Bonissone
Day 5: Adaptive Fuzzy and Neural Network Systems -- 
	Hamid Berenji

                        General Information

Registration Fee:

$1395	Includes tuition, a full copy of the notes, and 
        refreshments during the morning and afternoon breaks

$1295	For teams of three or more from the same organization.

A $75 processing fee is charged if registration is cancelled before July 2, 
1993.  No refund after July 2, 1993 but substitution is allowed at all 
times.

Location and Accommodation:

The course will be held at the beautiful hotel Sofitel.  Please 
arrange accommodation directly with the hotel.  Special rates are 
available by mentioning "IIS Corp. Fuzzy Logic Course".

Hotel Sofitel San Francisco Bay
223 Twin Dolphine Drive
Redwood City, CA 94065
Telephone: (415) 598-9000 X-6201

For further info, please contact:
              Intelligent Inference Systems Corp.
              P.O. Box 2908
              Sunnyvale, CA 94087
              Fax: (408) 730-8550
              email: iiscorp@netcom.com

------------------------------

Date: Tue, 22 Jun 93 11:48:32 BST
From: C A Gill <sta6cag@leeds.ac.uk>
Subject: Workshop: Shape Recognition in 2-D and 3-D Images

    FINAL NOTICE     FINAL NOTICE     FINAL NOTICE     FINAL NOTICE   
               13TH LEEDS STATISTICAL RESEARCH WORKSHOP               
                         JULY  12 - 14,  1993                         

                SHAPE RECOGNITION IN 2-D AND 3-D IMAGES
KEYNOTE LECTURES
Professor Josef Kittler, University of  Surrey,  UK,  will  present  a 
paper on  parametric  shape  detection  based  on  robust  statistical 
hypothesis testing with applications  in  computer  vision  and  image 
analysis.
Professor Michael Miller, Washington University, Missouri,  USA,  will 
give two talks on jump-diffusion processes for inferring 1, 2, and 3-D 
shape as used in electron microscopy and automatic target tracking and 
recognition.
OTHER CONTRIBUTIONS
Dr Tim Cootes,        Wolfson Image Analysis Unit, Manchester University
Dr Ian Dryden,                              Statistics, Leeds University
Dr Nick Efford,                       Computer Studies, Leeds University
Professor Colin Goodall,                     Statistics, Penn State, USA
Dr Simon Hesketh,                Wheatstone Labs., King's College London
Professor David Hogg,                 Computer Studies, Leeds University
Dr Andreas Lanitis,   Wolfson Image Analysis Unit, Manchester University
Professor John Marchant,                                      SILSOE, UK
Professor Kanti Mardia,                     Statistics, Leeds University
Dr Steve Maybank,                                             G.E.C., UK
Professor Michael Smith,    Centre for Medical Imaging, Leeds University
Professor Michael Werman,         Computer Science, Hebrew University of
                                                       Jerusalem, Israel
            POSTER DISPLAY          HANDS-ON SUN WORKSTATIONS           
TIMETABLE   Monday 1.00 p.m.(registration)  -  Wednesday 4.00 p.m.

REGISTRATION FEE : 78 pounds (students 58 pounds)
The workshop will be held in Fairbairn House, part of the University  of 
Leeds campus, and which offers conference and associated facilities to a 
high standard.  Fees include  lunch  on  Tuesday  and  Wednesday,  light 
refreshments,car parking and course material.
CONFERENCE DINNER (Tuesday): 12.50 pounds (payable with registration fee).
Fees should be paid in advance, cheques payable to UNIVERSITY OF LEEDS.
ACCOMMODATION in nearby hotels at  reasonable  prices  can  be  arranged 
if required.  Individuals must settle their own accommodation expenses.

If you  would  like  to  attend  the  workshop  then  for  more  details 
contact                    Dr. Christine Gill, Department of Statistics,
                                     University of Leeds, Leeds LS2 9JT.
telephone - Leeds (0532) 335157          email - sta6cag@uk.ac.leeds.gps

                 CLOSING DATE FOR BOOKINGS  -  FRIDAY  25TH  JUNE 1993

       SHAPE RECOGNITION IN 2-D AND 3-D IMAGES  12 - 14 JULY 1993       
REGISTRATION FORM Please detach and return to Dr C. A. Gill,  Department 
of Statistics, University of Leeds, Leeds LS2 9JT.

     NAME:...............................     Fee enclosed(pounds)
  ADDRESS:...............................          78 or 58   ....
          ...............................         Dinner 12.50....
          ...............................                TOTAL....
Please tick if you require:-
conference dinner......    vegetarian menu......
car park space......   sandwich lunch on Monday......
If accommodation required, please indicate
Monday night....  Tuesday night....  en suite....  price range......

------------------------------

End of VISION-LIST digest 12.28
************************
