Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site utcsrgv.UUCP
Path: utzoo!utcsrgv!phyllis
From: phyllis@utcsrgv.UUCP (Phyllis Eve Bregman)
Newsgroups: ont.events
Subject: J. Crowley:  "Multiple resolution representation and matching of gray scale shape".
Message-ID: <3756@utcsrgv.UUCP>
Date: Thu, 5-Apr-84 18:39:48 EST
Article-I.D.: utcsrgv.3756
Posted: Thu Apr  5 18:39:48 1984
Date-Received: Thu, 5-Apr-84 20:17:26 EST
Organization: CSRG, University of Toronto
Lines: 33

****
      UofT Department of Computer Science Seminar Schedule for
		the week of April 9th, 1984

Tuesday, April 10th, 2:00 P.M., GB244:  Professor James Crowley,
   CMU, Robotics Institute, Carnegie-Mellon University, Pittsburgh,
   PA:  "Multiple resolution representation and matching of gray
   scale shape".

   ABSTRACT:  This talk introduces a multiple resolution representation
   for 2-D gray scale forms, and describes algorithms for learning and
   matching 2-D pattern models based on this representation.  The
   representation is based on a reversible transform, called the
   Difference of Low Pass (DOLP) transform, which converts an image
   into a sequence of band-pass images.  The DOLP transform will be
   defined and a fast O(N) algorithm will be briefly described for
   its computation.

   A representation based on "peaks" and "ridges" in the DOLP transform
   space will then be introduced.  This representation expresses gray
   scale forms as a graph structure of symbols with attributes.  A
   description technique for gray scale forms can be defined using this
   representation which can be matched invariant to size, orientation,
   position or intensity.  A training algorithm will be described
   for learning structural models of 2-D gray scale patterns.  A
   structural matching algorithm based on these models will then be
   described.
-- 
		Phyllis Eve Bregman
		CSRG, Univ. of Toronto
		{decvax,linus,ihnp4,uw-beaver,allegra,utzoo}!utcsrgv!phyllis
		CSNET:  phyllis@toronto
		(416) 978 6985