Copy Link
Add to Bookmark
Report
VISION-LIST Digest Volume 11 Issue 07
VISION-LIST Digest Thu Feb 13 10:21:54 PDT 92 Volume 11 : Issue 7
- 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:
Non deterministic Relaxation for Perceptual Grouping (Query)
Need fast region growing algorithm
Source code wanted for a binary image pre-processing algorithm.
Stereo equipment information wanted
Image acquisition board for EISA bus
Page-scanned Brodatz book?
Parallel computers
Summary of 3-D object generation packages (long)
Summary: Principal axes/axes of symmetry (long)
Position: Permanent Lectureship in Philosophy, University of Sussex
Faculty Position in Radiological Computing
New Vision Book
_EJournal_
----------------------------------------------------------------------
Date: Wed, 12 Feb 92 15:58:15 GMT
From: G.Nicholls@ee.surrey.ac.uk
Subject: Non deterministic Relaxation for Perceptual Grouping (Query)
Hi there,
I'd like to hear of any work on stochastic approaches to high
level vision and in particular, non-deterministic relaxation (NDR) methods
for doing perceptual grouping. Is anyone out there working in this
area ?
We have been investigating a multi-resolution application of the
metropolis algorithm and have learnt a little about transforming model
knowledge (in the form of a Gibbs Potential for an MRF) from fine to
coarse resolution. This is only a small step up the image-descriptor
heirachy, but we want to know if its practicable to take steps higher
up the ladder in an NDR-MRF framework.
I will sumarise responses to the net.
Geoff Nicholls, Elec Eng, Uni of Surrey, UK.
------------------------------
Date: Fri, 7 Feb 92 23:03:04 -0500
From: harsh@eceris.ece.ncsu.edu (Harsh Potlapalli)
Subject: Need Fast Region Growing Algorithm
We are attempting a comparison of image segmentation techniques. To
increase our throughput we need a fast region growing algorithm. Any
help in locating public domain software and/or literature for this task
is greatly appreciated.
Thank You
-harsh
(harsh@eceris.ece.ncsu.edu)
[ Fastest one I know of is constant time. It uses city block distances
computed by a chamfer algorithm (two passes, six image positions) plus
a threshold on the amount you wish to grow. phil... ]
------------------------------
Date: 11 Feb 92 13:58:51 GMT
From: Claude-Marie Lafforgue <claude@sunlite.concordia.ca>
Organization: Concordia University, Montreal Quebec
Subject: Source code wanted for a binary image pre-processing algorithm.
I'm searching for a good preprocessing algorithm to do some smoothing
and noise elimination for binary images of characters for use in my
character recognition research.
Claude-Marie Lafforgue
CENPARMI (Centre for Pattern Recognition and Machine Intellignce)
Concordia University
------------------------------
Date: Fri, 7 Feb 92 19:21:58 GMT
From: martin@eola.cs.ucf.edu (Glenn Martin)
Subject: Stereo equipment information wanted
We are interested in buying a stereo system for digitizing a pair of
images. The system should be flexible so that positioning and
orientation of cameras are possible. Also any information regarding
the useful equipment for calibration, lenses, truntables, etc. useful
for vision research in stereo and motion will be helpful. Are there
any vendors selling this kind of equipment?
Thanks ...
Glenn Martin
------------------------------
Date: 11 Feb 92 19:38:20 GMT
From: rlp@bsu-cs.bsu.edu (Ralph L. Place)
Organization: CS Dept, Ball St U, Muncie, Indiana
Subject: Image acquisition board for EISA bus
I have been using a Truevision M8 board in a Zenith PC for image
acquisition in a robot vision setup and have recently gotten a Hewlett
Packard 486 machine with an EISA bus. I am looking for info about an
EISA monochrome image acquisition/display board that will work with
the HP computer. Does anyone have any info on such a board?
R. Place
Dept of CS
BSU
Muncie, IN 47306
rlp@bsu-cs.bsu.edu
------------------------------
Date: Fri, 07 Feb 92 15:54:06 EST
From: John_Krumm@IUS5.IUS.CS.CMU.EDU
Subject: Page-scanned Brodatz book?
Has anyone else used a page scanner to digitize the Brodatz texture book?
I've just assembled all the hardware and software I need to scan in
the Brodatz book at 300 dpi on a page scanner and get the images in a
format I can use. Since the images in the book are about 9.5x7.5
inches, this will give an incredible resolution of 2850x2250 pixels.
(I found that scanning at 100 dpi caused moire patterns. I suppose
the scanning frequency was too low compared to the dot pattern
resolution of the printing.)
We've got two digitized versions of the Brodatz book here, and they're
both lacking in some respects. I suspect they were digitized using a
video camera and digitizer. (Interestingly, although their resolution
is about the same as that which caused moire patterns for me, they
don't have moire patterns. Inadvertent lens defocus must have taken
care of that problem. This also leads me to believe that our page
scanner is well-focussed.)
Before I do the scanning, I want to know if anyone else has already
done something like this. If so, and if I can get the images from
you, I won't bother.
John Krumm
------------------------------
Date: 12 Feb 92 15:26:26 GMT
From: gregoire@bora.inria.fr (Gregoire Malandain)
Organization: INRIA
Subject: parallel computers
Hi,
I am working with 3D images, and some of the algorithms I used are easily
parallelizable. Then I am looking for informations on the use of
parallel computers in image analysis. Does anyone of you use parallel
computers ? For what kind of applications ? Any general information on
this subject will be wellcome. More specifically, I am looking for
informations on KSR computers.
thanks for any help you can provide
Gregoire
email : Gregoire.Malandain@inria.fr
------------------------------
Date: Tue, 11 Feb 92 21:53:19 GMT
From: rkc@xn.ll.mit.edu
Organization: MIT Lincoln Laboratory
Subject: Summary of 3-D object generation packages
Many kind fellow researchers responded to my request for a program that would
enable me to create and manipulate 3D data. Many more wrote to request copies
of the responses I received. I present the following summary so others need
not work so hard to find this information:
Suggestions came in several forms. The hardest to make work (in my opinion)
were drawing packages in which one creates stacked 2D contours to create 3D
objects, and then translate these around. This was not quite what _I_ wanted,
but others may find this useful. Here are some of these packages:
From: Amir A. Amini <amini@noodle.med.yale.edu>
>The package is called "tgif" and from my experience, is good for drawing 2D
>contours which can be stacked on top of one another into making a 3D object.
>The authors name is William Cheng (william@cs.ucla.edu).
From: Amir A. Amini <amini@noodle.med.yale.edu>, also noted by
From: sundare@SPUNKY.CS.NYU.EDU (Venkataraman Sundareswaran)
>Another piece of software which might be helpful ( I am not sure if it is
>public domain ) is Sandy Pentland's "THINGWORLD" which uses superquadrics as
>the 3D primitive. His e-mail address is sandy@media-lab.media.mit.edu
One step better than the 2-D draw programs were the 3-D CAD programs.
dandrea@prsun02.cineca.it (Vincenzo D'Andrea) had this suggestion:
>I've heard of BRL-CAD, according to my information is just what you
>need. You can download it from trident.arc.nasa.gov (128.102.18.13).
Others suggested I use a ray tracing program, and a series of image
manipulation programs. One person (richards@eleceng.ee.queensu.ca (Haydn
Richardson)) suggested I try a package called RADIANCE. Furthermore, Haydn
said:
>RADIANCE is available from Greg Ward at Berkeley. You can
>get RADIANCE by anonymous ftp to hobbes.lbl.gov or you can contact Greg at
>greg@hobbes.lbl.gov. RADIANCE probably has all the modelling abilities you
>need.
>I wrote a simple program to convert a RADIANCE scene model file into a
>sequence of model files where each object has its own set of motion
>parameters. I used the 3-d motion model presented by Young and Chellapa in
>their August 1990 paper in the IEEE Transactions on Pattern Analysis and
>Machine Intelligence. I used MATLAB to do the 3-d matrix arithmetic and
>RADIANCE to render the individual images.
One person (<stricker@cs.uchicago.edu>) suggested Khoros. Acording to Markus:
> The image processing package Khoros (it is free) contains an
>easy interface to generate synthetical images (like squares, spheres,
>..). In addition, they provide a lot of tools to fiddle around with
>the images.
>Khoros is free but huge (60 Meg). Runs under X on Suns and SGI's. I am
>sure they have it somewhere at MIT otherwise you can ftp it from U of
>New Mexico.
Several people suggested I try out "rayshade", whose authors include:
> Craig Kolb Rod Bogart
> rayshade@weedeater.math.yale.edu
I ftp'd this program because it has some built-in animation (translation and
rotation about an arbitrary point--tricky to use, but it gets the job done),
light source specification, shadows, fog, and stereo rendering. This seems to
be quite useful, but really I haven't used it enough yet to comment on its
stability. I do note that the package is in release 4.0, so presumably it has
been around for awhile.
Best of luck,
-Rob
------------------------------
Date: Sun, 9 Feb 92 22:45:17 EST
From: raja@cps.msu.edu
Subject: Summary: Principal axes/axes of symmetry
Here is a summary of reponses, as requested by several people. Thanks
to those who responded!
Regards,
Narayan S. Raja.
From: kcho@occlusal.rutgers.edu (Kyugon Cho)
Subject: Detection of symmetries.
The first preprocessing would be the binarization of the images.
Next, you can try one of the following techniques depending on your purpose of
finding the symmetries.
1. Thinning algorithm.
2. Morphological skeletonization.
3. Find the boundary of the objects and apply
SAT, SLS, or HLS.
From: kcho@occlusal.rutgers.edu (Kyugon Cho)
Subject: Re: Detection of symmetries.
The following papers are not good introductory papers about thinning and
morphological thinning but they are quite recent.
You could find good indirect references.
@Article{jang90,
author = "B.~Jang and R.~T.~Chin",
title = "Analysis of thinning algorithms using mathematical
morphology",
journal = PAMI,
year = "1990",
volume = "12",
number = "6",
pages = "541-551",
OPTmonth = "",
OPTnote = ""
}
@Article{mahmoud91,
author = "S.~A.~Mahmoud I.~Abuhaiba and R.~J.~Green",
title = "Skeletonization of {Arabic} characters using
clustering based skeletonization algorithm {(CBSA)}",
journal = PR,
year = "1991",
volume = "24",
number = "5",
pages = "453-464",
OPTmonth = "",
OPTnote = "Output is a set of straight line segments rather than
points. Insensative to noise."
}
The SLS is "smoothed local symmetries" by M. Brady.
The HLS is "hierarchical local symmetries" by me and S. Dunn.
My paper appeared in the "Pattern Recognition Letters", June 1991.
It contains references about SAT, SLS, and other similar approaches.
- Kyugon.
From: Lance Gatrell (303) 977-2052 <gatrell@saturn.den.mmc.com>
Subject: Software to detect symmetry/principal axes?
I believe that Rod Grupen, Chuck Hansen, and Tom Henderson wrote some
software at the University of Utah to do this from range data. This
might be of use to you. They published a technical report on it.
The last that I heard (its been over a year), Rod was at UMass at Amherst
(email grupen@cs.umass.edu), Tom is the department chairman at the University
of Utah (email tch@cs.utah.edu).
From: Patrick J. Flynn <flynn@eecs.wsu.edu>
Subject: Re: Principal axes/axes of symmetry
> I was wondering if you knew of any work on
> finding the ``right'' axis of symmetry or
> principal axis of simple parts? For example,
> one can estimate axes from the eigenvectors,
> but which of them is the ``right'' principal
> axis? It is not always the one along which
> the object is most elongated (e.g. consider a
> flattened cylinder). If you know any relevant
> references, pl. let me know...
The eigenvector method you mention for axis finding has a more serious problem;
it is not invariant to occlusion. One invariant technique is to project
estimated surface normals into the object of interest. They should all
intersect the axis of symmetry if there is one. I asked Tim Newman to
look at that idea last year, but I don't think he had a lot of success
with it.
In the intensity image domain, tangent vectors taken from corresponding
points oin the silhouette will always intersect along the axis of symmetry
(actually the axis' projection in 2D). The has been used by a few people.
I think Vic Nalwa had a paper in CVPR '88 or '89 about it, and a paper in
PAMI within the last few years. Also, I think Kashi Rao (Ram Nevatia's
student) was doing something with symmetry in intensity data. Stockman
is aware of his work, perhaps you could touch base with him about it.
There has been a recent burst of interest in finding the cross-sections
of generalized cylinders that may have some bearing on your problem; check
out work by Ari Gross (Boult's student) and Jean Ponce
in recent CVPR/ICCV proceedings.
That's all I can think of at present. I hope it proves helpful.
Pat
From: newman@cps.msu.edu (Timothy Newman)
Subject: axis of symmetry
I think that Yokoda and Levine have some papers on recovering
axes of solids of revolution. Might be useful to your needs
for geons/generalized cylinders. They had a paper in ICPR 90 which
will have some references.
------------------------------
Date: Fri, 7 Feb 92 18:28:46 GMT
From: Andy Clark <andycl@syma.sussex.ac.uk>
Subject: Position: Permanent Lectureship in Philosophy, University of Sussex
**** PERMANENT LECTURESHIP IN PHILOSOPHY, UNIVERSITY OF SUSSEX ****
Applications are invited from men and women for a permanent lectureship in
philosophy at the University of Sussex, England. The applicant should be able
to teach "core" philosophy courses (e.g. Descartes to Hume), and should have a
specialist interest in the philosophy of mind/cognitive science and/or the
philosophy of language.
Application-forms and Further Particulars can be obtained from:
Ms. Elinor Mitchenall, The Personnel Office, Sussex House,
University of Sussex, Falmer, Brighton, BN1 9RH
Phone: (0273)- 678201
FAX; (0273) - 678335 (mark FAX "for the attention of the Personnel Dept."
------------------------------
Date: Tue, 11 Feb 92 19:33:09 EST
From: dsc@stapes.ent.hmc.psu.edu (david s. channin)
Subject: Faculty Position in Radiological Computing
Penn State University
The Milton S. Hershey Medical Center
Department of Radiology
Chief, Section of Radiologic Computing and Imaging Science
The Penn State University, Department of Radiology is seeking applications
for a new non-clinical faculty position in radiological computing and imaging
science. The responsibilities of this position include:
Performing original research in radiological computing
e.g., 2D and 3D radiologic image processing and analysis,
picture archive and communication systems (PACS),
and computer assisted roentgenological interpretation.
Planning the future development of the PACS network and computer
applications.
Representing the department in communications with medical center and
university computing interests.
Coordinating computer education projects for faculty, staff,
residents and students in the department.
Qualifications for the position include a Ph.D. or M.D./Ph.D. in
computer science or related fields along with several years experience
in large software and/or hardware project design and management.
Proven ability to secure extramural funding, appropriate publication
record, and excellent technical skills including knowledge of UNIX,
the C language, relational database systems, networks and modern
computing paradigms are important prerequisites, along with superior
interpersonal skills.
Computing resources in the Department of Radiology include a
3D imaging workstation, a Sun computer network, a Macintosh network
and access to an IBM ES390-9000 mainframe with a vector facility and
connection to the internet through Penn State University.
Penn State University Hospital is a 500 bed tertiary care facility
with a level I trauma center, located in Hershey, Pennsylvania, ten
miles east of Harrisburg, the state capital. Hershey is 2 hours by car
from Washington, D.C. and Baltimore and three hours from New York City.
An excellent benefits package is provided, and the level of faculty
appointment and salary will be commensurate with the applicant's
experience and training.
Applicants should respond with a letter of interest, current curriculum
vitae and list of three references to:
John E. Madewell, M.D.
Professor and Chairman
Department of Radiology
Penn State University
The Milton S. Hershey Medical Center
P.O. Box 850
Hershey, PA 17033
Penn State is an affirmative action, equal opportunity employer.
Women and minorities are encouraged to apply.
------------------------------
Date: Tue, 11 Feb 92 15:36:24 GMT
From: David Wilson <david@turing-institute.ac.uk>
Subject: New Vision Book
The following book has recently appeared and may be of interest:
Michael S. Landy & J. Anthony Movshon (Eds.)
Computational Models of Visual Processing
MIT Press
Contents:
THE TASK OF VISION
Adelson & Bergen - The Plenoptic Function and the Elements of Early Vision
RECEPTORS AND SAMPLING
Ahumada - Learning Receptor Positions
Tiana, Williams, Coletta and Haake - A Model of Aliasing in Extrafoveal Human
Vision
Hood and Birch - Models of Human Rod Receptors and the ERG
MODELS OF NEURAL FUNCTION
Lennie, Haake & Williams - The Construction of Chromatically Opponent Receptive
Fields
Hawken & Parker - Spatial Receptive Field Organization in Monkey V1 and its
Relationship to the Cone Mosaic
Watson - Neural Contrast Sensitivity
Shapley and Reid - Spatiotemporal Receptive Fields and Direction Selectivity
Heeger - Nonlinear Model of Neural Responses in Cat Visual Cortex
DETECTION AND DISCRIMINATION
Nachmias - A Template Matching Model of Subthreshold Summation
Pelli - Noise in the Visual System May be Early
Wilson - Pattern Discrimination, Visual Filters, and Spatial Sampling
Irregularity
COLOR AND SHADING
Brainard & Wandell - A Bilinear Model of the Illuminant's Effect on Color
Appearance
D'Zmura - Shading Ambiguity: Reflectance and Illumination
Kersten - Transparency and the Cooperative Computation of Scene Attributes
MOTION AND TEXTURE
Grzywacz & Yuille - Theories for the Visual Perception of Local Velocity and
Coherent Motion
Bergen & Landy - Computational Modeling of Visual Texture Segregation
Graham - Complex Channels, Early Local Nonlinearities, and Normalization in
Texture Segregation
Chubb & Landy - Orthogonal Distribution Analysis: A New Approach to the Study
of Texture Perception
3D SHAPE
Bulthoff - Shape From X: Psychophysics and Computation
Frisby & Pollard - Computational Issues in Solving the Stereo Correspondence
Problem
Parker, Johnston, Mansfield & Yang - Stereo, Surfaces and Shape
------------------------------
Date: Tue, 11 Feb 92 23:07:23 GMT
From: ejournal@csc.albany.edu (Edward M. Jennings)
Organization: State University of New York at Albany
Subject: _EJournal_
Announcing an Electronic Journal February 1992
_EJournal_ is a peer-reviewed, all-electronic, network distributed,
serial publication.
We are particularly interested in theory and practice surrounding the
creation, transmission, storage, interpretation, alteration and
replication of electronic "text," broadly defined.
We are also interested in the social, psychological, literary, economic,
pedagogical, philosophical and other ramifications of computer-mediated
networks.
Our review process is anonymous, all-electronic, and consensual.
We prefer brief, authentic, lively essays to exhaustive technical reports.
Single-essay issues appear as often as submissions are affirmatively
reviewed; there were four (free) distributions to subscribers in 1991.
There are two dozen consulting editors, in several disciplines, who
review submissions. Members of _EJournal_'s advisory board are:
Stevan Harnad, Princeton University
Dick Lanham, University of California at Los Angeles
Ann Okerson, Association of Research Libraries
Joe Raben, City University of New York
Bob Scholes, Brown University
Harry Whitaker, University of Quebec at Montreal
To subscribe to _EJournal_, send a mail message to listserv@albnyvm1.bitnet
containing as its only line the command:
subscribe ejrnl your_first_name your_last_name
Information about getting back issues will accompany the "Welcome"
message sent to people who subscribe.
Please send submissions for editorial consideration to our "office" at:
ejournal@albnyvms.bitnet
Ted Jennings, Editor, Department of English, University at Albany/SUNY
------------------------------
End of VISION-LIST digest 11.7
************************
