<p>I am using ENVI to do something relatively simple: I have taken photos of plants and want to categorize the different colors to count the number of pixels in each (i.e. green=living, dead=brown, background=white). The photos are tracking the same plants over time and have been taken in the same place against a white background. I have done unsupervised classification (kmeans) and supervised classification (maximum likelihood). The maximum likelihood produced the results I was looking for, however I would have to create training samples for each photo and I already have 1000 photos that need to be analyzed. Does anyone know of a way to get this result with a method that could categorize the pixels for many photos without having to categorize each one?</p>