Newsgroups: sci.image.processing,comp.graphics.algorithms
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!newshost.marcam.com!zip.eecs.umich.edu!newsxfer.itd.umich.edu!agate!howland.reston.ans.net!torn!news.bc.net!newsserver.sfu.ca!fornax!manderso
From: manderso@cs.sfu.ca (Mark Anderson)
Subject: Re: Connected components routines for greyscale iamges?
Message-ID: <524*manderso@cs.sfu.ca>
Reply-To: manderso@cs.sfu.ca (Mark Anderson)
Organization: SFU/TRIUMF Medical Computing Laboratory
References: <3hauk1$r5l@mozo.cc.purdue.edu> <519*manderso@cs.sfu.ca> <3hf3aj$djv@mozo.cc.purdue.edu>
Date: Sat, 11 Feb 1995 18:21:17 GMT
X-Url: http://fas.sfu.ca/1/cs/people/GradStudents/manderso
Lines: 17
Xref: glinda.oz.cs.cmu.edu sci.image.processing:12589 comp.graphics.algorithms:13187

jonesbr@cartoon.ecn.purdue.edu (Brian R. Jones) writes:
>The only problem is: how does one decide how
>much of the peak in the histogram corresponds to an object?  Looking
>at a specific image and the corresponding histogram, I could make this
>judgement, but how do I get the computer to do this so that it will
>work relably on images of different compositions, brightnesses and
>contrasts?  An arbitray selection of say, 80% of the peak, may be
>fine for one image, but horribly inaccurate for another.

You might want to try "moment-preserving threshold"; we have used this
technique to automatically threshold greyscale images. I'm not too 
familiar with how it works, but it doesn't require you to choose a
threshold. Here's the reference:
W. Tsai, "Moment-Preserving thresholding: A new approach", Computer
Vision, Graphics and Image Processing, Vol. 29, 1985, pp 377-393.
-- 
mark
