>
> Start with a nice, high contrast image (because thresholding and
> segmentation will be needed).  Then, simply threshold the image to an
> appropriate level that defines the "clumps" you are looking at.  From here,

You will find that intensity based segmentation will often work much
better if you "flatten" the background. Even better  is to do
a true background correction. However, if you already have your images,
and you lack true background images, you can do pretty well by using a
very course low pass filter (something that averages over an area larger than
your objects) and then subtracting that from the original.

Image Pro Plus (which I use a lot) has background extraction and flatten
functions built in, and all the other packages can also do some variation
of this.

--aryeh
Aryeh Weiss                          | email: [log in to unmask]
Department of Electronics            | URL:   http://optics.jct.ac.il/~aryeh
Jerusalem College of Technology      | phone: 972-2-6751146
POB 16031                            | FAX:   972-2-6751275
Jerusalem, Israel                    | ham radio: 4X1PB/KA1PB