Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site octopus.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxt!houxm!hjuxa!petsd!pesnta!pyramid!octopus!pete
From: pete@octopus.UUCP (Pete Holzmann)
Newsgroups: net.math,net.graphics
Subject: Looking for 2D point fitting alg's: Rubber Sheet or BiLinear
Message-ID: <191@octopus.UUCP>
Date: Fri, 24-Jan-86 17:09:32 EST
Article-I.D.: octopus.191
Posted: Fri Jan 24 17:09:32 1986
Date-Received: Sat, 25-Jan-86 09:37:20 EST
Reply-To: pete@octopus.UUCP (Pete Holzmann)
Organization: Octopus Enterprises, Cupertino, CA
Lines: 22
Xref: watmath net.math:2716 net.graphics:1414


For a digitizing application, I'm looking for any good algorithms that
can be used to generate a transformation function of (x,y) given a set
of reference tranformations.

E.g., I have N sets of (x,y) -> (x',y') values, and want a general way
to come up with F(x,y) that gives (x',y') for any (x,y).p

I've heard that applying a least squares fit to a polynomial in x,y,x^2,y^2 and
xy gives a good fit, but haven't been able to figure out how to
implement that in practical terms for my application: I keep getting
results that don't work at all.

Has anybody out there worked on problems like this? Any suggestions or
reference sources I can go to?

Thanks!
-- 
  OOO   __| ___      Peter Holzmann, Octopus Enterprises
 OOOOOOO___/ _______ USPS: 19611 La Mar Court, Cupertino, CA 95014
  OOOOO \___/        UUCP: {hplabs!hpdsd,pyramid}!octopus!pete
___| \_____          Phone: 408/996-7746