Newsgroups: sci.image.processing
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!europa.eng.gtefsd.com!gatech!psuvax1!ukma!rsg1.er.usgs.gov!speclab.cr.usgs.gov!ngorelic
From: ngorelic@speclab.cr.usgs.gov.cr.usgs.gov (Noel S. Gorelick)
Subject: [Q] Clustering Algorithms?
Message-ID: <1994Dec14.220958.5338@rsg1.er.usgs.gov>
Sender: news@rsg1.er.usgs.gov
Reply-To: ngorelic@speclab.cr.usgs.gov
Organization: US. Geological Survey, Denver
X-Newsreader: TIN [version 1.1 PL6]
Date: Wed, 14 Dec 1994 22:09:58 GMT
Lines: 23

Trying to find any information on cluster detection techniques.  

We have images that are essentially abundance maps, ie: data value
of each pixel is proportional to the abundance of some substance,
and a value of zero everywhere the substance is not found.  (maps
are mostly zeros)

What we hope to see are areas (clumps) of data values, on an
otherwise black background, but quite often, due to noise in the
detection system, we get static, spread across the image, which inflates
the 'number of pixels found'.

We would like to be able to determine statistically, how 'clumpy'
the image is.  And perhaps some way to cluster together pixels that
appear statistically related.

Any help appreciated, e-mail, please

Noel (ngorelic@speclab.cr.usgs.gov)
--
   Noel Gorelick - MercWorks Software and Digital Solutions, Denver
   ngorelic@m-net.arbornet.org // ngorelic@speclab.cr.usgs.gov
   "Quis custodiet ipsos custodes -- Who watches the watchmen?"
