MATLAB: Which technique is best for classifying using color variation

Question

I have to classify the thermal image without any thermal data in it. It is basically similar to classifying objects. I want to know which classifier is best for my case. There will be only color variations in the image with change in temperature. I tried using Bag of Features concept but My accuracy is only 50%. Which interest point detectors I can use. Please help me out. Thanks

Answer 1

If you are still working on the same project as before:

Average the "normal" images to create a base image. Calculate the maximum MSE (Mean Squared Error) of the "normal" images relative to the base image. That gives you an upper bound on normal images.

Now your classification is just whether MSE of the current image compared to the base image is greater than the threshold or not.