How to apply unsupervised classification to spatial data

Question

I am trying to learn how to apply unsupervised classification to spatial data. In near infra red satelite pictures; the ocean is dark and the forest is white. Each pixel in such an image is an observation. I guess the variables are [x-position, y-position, grayscale]?

Which methods could I use to classify the pixels into ocean and forest? I want to ensure that the points in the classes forms spatially connected regions (e.g. no single trees in the ocean).

Thanks in advance for any answers!

Answer 1

This sounds to me like an image processing question, unless you are looking for a very complex structure. You may want to use a gaussian filter on the image, and then apply a threshold. Also, you can ask in http://dsp.stackexchange.com/ .

Answer 2

This software (it won the best demonstration award at SSTD 2011) http://www.springerlink.com/content/2x3258q4241j6602/ should be able to do spatial clustering, too.