Copy Link
Add to Bookmark
Report
VISION-LIST Digest 1989 06 16
Vision-List Digest Fri Jun 16 09:19:51 PDT 89
- Send submissions to Vision-List@ADS.COM
- Send requests for list membership to Vision-List-Request@ADS.COM
Today's Topics:
rule based decisions vs functional decisions
Image Repository
Range Images
RE: Vision-List delayed redistribution
Range imagery
Re: Request for range images
Neural nets for vision
Where is Alv software ?
Vision/Image Processing Languages
video recording equipment
----------------------------------------------------------------------
Date: 3 Jun 89 01:47:52 GMT
From: Phillip Neal <nealiphc@BLAKE.ACS.WASHINGTON.EDU>
Subject: rule based decisions vs functional decisions
Keywords: image understanding, machine learning , image processing
Organization: Univ of Washington, Seattle
Does anybody have any performance numbers for a rule based decision
making image segmentation method vs a regular functional decision
making method.
In other words, which is better -- theoretically or empirically in terms
of classification rate:
1. Developing rules like:
If the edge is more than 20 grey levels and the major direction is north
east, then the confidence factor for this rule is 20
else
confidence factor for this rule is 0
vs
2. y = f(delta(grey),direction)
if(y.gt..5) then this is a true edge.
and the coefficients are 'learned' through some inductive process
like discriminant analysis or some bayesian update routine.
That's all for now,
Phil Neal ---- nealiphc@blake.acs.washington.edu
------------------------------
Date: Mon, 5 Jun 89 09:08:43 BST
From: alien%VULCAN.ESE.ESSEX.AC.UK@CUNYVM.CUNY.EDU (Adrian F. Clark)
Subject: Image Repository
I'm interested in collecting together a set of `standard' images of
all types--TV, remote sensing, tomography, range, etc to test out a
package I'm developing. I'm also looking at ways of making imagery
available over a network (ie, lossless coding into that ASCII subset
which will pass through most gateways unscathed). The sort of imagery
I'm looking for should be both good (standard algorithms work) and bad
(standard algorithms fail). If anyone else is interested in such an
idea, please contact the author...especially if you've some suitable
imagery!
Adrian F. Clark
JANET: alien@uk.ac.essex.ese
ARPA: alien%uk.ac.essex.ese@nss.cs.ucl.ac.uk
BITNET: alien%uk.ac.essex.ese@ac.uk
Smail: Dept. of Electronic Systems Engineering, Essex University,
Wivenhoe Park, Colchester, Essex C04 3SQ, U. K.
Phone: (+44) 206-872432 (direct)
"The great tragedy of Science--the slaying of a beautiful
hypothesis by an ugly fact." -- T H Huxley (1825-95)
------------------------------
Subject: Range Images
Date: Mon, 05 Jun 89 10:06:58 -0400
From: "Kenneth I. Laws" <klaws@nsf.GOV>
One source of range images is the NRCC Three-Dimensional
Image Data Files. The collection includes: simple objects
with planar, spherical, and cylindrical surfaces, sometimes
overlapping; multiple views of isolated objects; human faces;
and various complex objects. For information about tapes
and diskettes (about $350 for each section of 40-66 images
on tape, $750 on 5.25" diskettes) contact
M. Rioux or L. Cournoyer
Photonics and Sensors Section
Laboratory for Intelligent Systems
Division of Electrical Engineering
National Research Council of Canada
Ottawa, Ontario, Canada K1A 0R8
(613) 993-7902
-- Ken Laws
National Science Foundation
------------------------------
Date: Mon, 5 Jun 89 10:46 CDT
From: "H. Ogmen, OGMEN@UHVAX1.UH.EDU, OGMEN@UHVAX1.BITNET"
Subject: RE: Vision-List delayed redistribution
Re: Three-dimensional image data files.
The National Research Council of Canada (NRC) has a large number of
three-dimensional image data files.
For information about these images (and prices) contact:
M. Rioux, L. Cournoyer
Photonics and Sensors Section
Laboratory for Intelligent Systems
Division of Electrical Engineering
National Research Council of Canada
Ottawa, Ontario, Canada
K1A 0R8
Tel. (613) 993 7902
Telex: 053-4134
Telefax: 613- 952-7998
H. Ogmen
Dept. of EE
University of Houston
------------------------------
Date: Mon 5 Jun 89 10:50:53-PDT
From: Gerard Medioni <MEDIONI%DWORKIN.usc.edu@oberon.usc.edu>
Subject: Range imagery
I am aware of at least two sets of "standard" range image databases,
one from university of Utah (Prof Thom Henderson), and one from the
National Research Council of Canada (M. Rioux and L. Cournoyer). The
first is available at a nominal fee, the second for a few hundred $$.
------------------------------
Date: Tue, 6 Jun 89 09:31:20 PDT
From: Bruce Bon <bon@saavik.Jpl.Nasa.Gov>
Subject: Re: Request for range images
There is an extensive set (214 images) of range images published by the
National Research Council Canada. The range data was taken with a
synchronized laser scanner. A book, "The NRCC Three-dimensional Image
Data Files," CNRC 29077, contains all of these images and is available
at no charge from:
Editorial Office, Room 301
Division of Electrical Engineering
National Research Council of Canada
Ottawa, Ontario, Canada
K1A 0R8
Machine-readable versions are available in several formats (ASCII/binary,
raw/interpolated) on 1600 bpi magtape and 5 1/4" DSDD diskettes. For
information, contact:
M. Rioux and L. Cournoyer
Photonics and Sensors Section
Laboratory for Intelligent Systems
Division of Electrical Engineering
National Research Council of Canada
Ottawa, Ontario, Canada
K1A 0R8
Telephone: (613) 993-7902
Telex: 053-4134
Telefax: (613) 952-7998
I hope this helps.
Bruce Bon
bon@robotics.jpl.nasa.gov (ARPAnet)
------------------------------
Date: 6 Jun 89 23:06:13 GMT
From: kroger@cs.utexas.edu (Jim Kroger)
Subject: Neural nets for vision
Keywords: neural nets, paten recognition, computer vision
Organization: U. Texas CS Dept., Austin, Texas
Are there any existing applications of neural net technology to object
recognition problems? I know that much theoretical work with neural
nets has involved vision, the work of David Marr being an example.
However, I am interested in creatng a system which can actually
recognize a sizeable number of objects. I am not sure whether this is
somehing that can now be accomplished in hardware, or must be
implemented in software. Can anybody advise me as to existing systems
or techniques, either hardware or software, which might accomplish
this task? Also, what kind of hardware is optimal for software
implementations? Basically, I want to know if neural net technology
offers any practical, workable solution to object recognition.
Any information or references will be greatly appreciated, and a
summary will be posted.
Jim Kroger
kroger@cs.utexas.edu
------------------------------
Date: Wed, 7 Jun 89 21:24:00 edt
From: parzen%jimmy@bess.HARVARD.EDU (Michael Parzen)
Subject: Where is Alv software ?
Back in November of 1988, I ftp'd some software written by Phill Everson
called Alv (autonomous land vehicle). It was some good vision software.
Does anyone know where I ftp'd this software from, i.e. where it can be
located on the net ? I forgot where it was and need to get it again.
Thanks in advance.
Mike Parzen
parzen@csc.brown.edu
------------------------------
Date: 10 Jun 89 00:13:20 GMT
From: mdavcr!rdr@uunet.UU.NET (Randolph Roesler)
Subject: Vision/Image Processing Languages
Summary: Wanted - Image Processing Language References
Keywords: Image Language
Organization: /etc/organization
I am looking for references to Image
processing languages. What I want is
a language which has builtin support
for high and low level image/vision
operations.
I.e. image -> image operations
image -> object operations
object -> object operations
Please, don't inform me of great general
purpose languages such as lisp, prolog, ....
Research and commercial languages OK.
We summarize to the Net.
Randy Roesler
MacDonald Dettwiler
Vancouver, BC, Canada
604-278-3411
uunet!van-bc!mdavcr!rdr
------------------------------
Date: 16 Jun 89 01:37:02 GMT
From: munnari!latcs1.oz.au!suter@uunet.UU.NET (David Suter)
Subject: video recording equipment
Organization: Comp Sci, La Trobe Uni, Australia
I have a query about the availabilty of video recording equipment
(manufacturers names if such exists) for a medical monitoring application.
Essentially the video signal has other information (other than the picture)
encoded during the blanking periods. The applications require to different
recording systems:
1. After a period of monitoring (say about 10 secs) the recording equipment
dumps 1 picture frame to tape or disc and also the average of some quantity
that has been encoded on the blanking periods.
2. After an event trigger a recording is made of THE 30 SECS TO 1 MIN
that occurred PRIOR to the Event - as well as 1 MIN or so after the event.
The general idea is that in both cases the encoded data as well as the
video data corresponding to this, can be analysed later to see what
was happening when interestin events occurred - in the later case it
is important to see what lead up to the event. The applications
concerned are not mine - so the details relayed may be a little hazy.
However, the general charcteristics required are as above. My contact
believes that there are commercial systems that do the type of thing
above (DATA VIDEO ENCODER?) but doesn't know sources of such
equipment. For 2. he wants any suggestions of ways of acheiving these
ends (somehow having the previous 1 min available to store if an
interesting event occurs). The system must be capable of running on a
video tape that is changed say every 3 hrs. Thus suggestion of
solutions he can get engineeered himself - or companies that deal in
equipment that provide this sort of functionality.
thanks.
d.s.
------------------------------
End of VISION-LIST
********************
