Xref: utzoo comp.sys.amiga.applications:912 comp.sys.amiga.graphics:1345
Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!wuarchive!psuvax1!atlantis.psu.edu!psuecl!rey
From: rey@ecl.psu.edu
Newsgroups: comp.sys.amiga.applications,comp.sys.amiga.graphics
Subject: Re: counting items in digitized images
Message-ID: <1991May8.005242.1172@ecl.psu.edu>
Date: 8 May 91 04:52:42 GMT
References: <1991May7.002855.4111@ducvax.auburn.edu> <0k7k22w161w@nstar.rn.com>
Organization: Engineering Computer Lab, Pennsylvania State University
Lines: 49

In article <0k7k22w161w@nstar.rn.com>, tbissett@nstar.rn.com (Travis Bissett) writes:
> bwaid@ducvax.auburn.edu writes:
>  
>> Does anyone know of a program that will count the number of occurences of
>> something in a digitized image. Some people at a medical school are trying to
>> count the number of cells in an image on an Amiga. Any help will be
>> appreciated.
>  
> Speaking from my own personal ignorance, I'd be surprised if you find a
> ready made canned solution to the problem you described. It really is
> non-trivial -- but then, it's the sort of challenge that many in academia
> and industry have been working on for a long time. In essence you've asked
> for a free and easy solution to the "robot vision" problem.  You want to
> take a digitized image, perform certain signal processing techniques, and
> end up with an unamgbiguous set of visual "objects" (whether they are tansk
> or cells or bolts in a bag). It certainly is in the realm of technical
> possibility to count cells from a microscopic camera picture -- I've yet to
> hear of a commercial explotation of it, though, which suggests that the
> problem is still in the experimental/developmental stages. The counting part
> is easy --once you get a clean image of idebtifying cell structure just
> adapt normal OCR algorithms. But getting a setup and program versatile
> enough to remove all variables of slide specimen, camera noise, etc. is a
> real task. Maybe a semi-automatic process is good enough for now, where some
> grad student makes the decision about what signal processing needs to be
> performed in order to get the acceptable level of image quality.
>  
> Sorry that I'm no help :-)
>  
> Travis
>  
> --
> Travis Bissett       NSTAR conferencing site       219-289-0287/317-251-7391
> internet: tbissett@nstar.rn.com              1300 newsgroups - 8 inbound lines
> uucp: ..!uunet!nstar.rn.com!tbissett            99 file areas - 4300 megabytes
> ---  backbone news & mail feeds available - contact larry@nstar.rn.com  ---
About 10 years ago, when I was a chemist working for Warner-Lambert, I recall
coming across such a device for colony counting.  It was part of an integrated
system in which the user slid a petri dish under a video camera.  The image of
the petri dish was displayed and digitized, meanwhile the computer would
encircle each colony and count it.  You realize of course, that the problem of
counting cultures is a very constrained problem making it relatively easy 
(culture's usually don't overlap and there is good contrast between the culture
and the medium).  The problem of cell counting in an overlapping situation is
indeed difficult, but I got the impression that the original poster was looking
for a simple colony counter.  Alas, I don't know where there is a colony counter
available for the Amiga, but if one turns up, I'd be interested in finding
out where you got it.

Bob