Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!seismo!rutgers!labrea!glacier!jbn
From: jbn@glacier.STANFORD.EDU (John B. Nagle)
Newsgroups: comp.ai,comp.graphics
Subject: Low-level feature extraction and interframe matching software wanted
Message-ID: <17150@glacier.STANFORD.EDU>
Date: Sat, 1-Aug-87 15:18:48 EDT
Article-I.D.: glacier.17150
Posted: Sat Aug  1 15:18:48 1987
Date-Received: Sun, 2-Aug-87 09:55:54 EDT
Organization: Stanford University
Lines: 20
Keywords: feature, image processing, computer vision, imaging
Xref: mnetor comp.ai:680 comp.graphics:992


     Does software exist for following moving objects from frame to frame
in video images?  I'm looking for something that works by finding low-level
features such as edges and corners and matches them from one frame to the
next.  I'm aware of "optical flow" calculation, but the usual numeric
method for doing this is differentiation-oriented and too noise-sensitive
to be useful on real-world images, I am told by someone who has tried it.
So I'd like something that finds many low-level features and tries to
match them up.

     The intended application is a vision system for a robot vehicle.  But it
is possible that techniques used for colorizing B/W films would be useful
for this purpose.  So I'd like to hear from people who know how colorizing
is done.  

     Software, algorithms, hardware, or indications of research activity
would be useful.

					John Nagle
					Center for Design Research, Stanford