I'm trying to think of a way to classify a set of observed pixel values based on prior knowledge.
I'm projecting an image with a set of known colored ordered vertical stripes (say a red stripe, a green stripe, magenta stripe and a black stripe in that order), on to some object (let the object be white, to make sure there is no change in color).
Now, I photograph the object with colors projected on top of it, and want to classify each observed pixel as one of the projected colors. Just to clarify if it isn't obvious, there will be noise in the photograph so magenta will never be magenta which I read values from the photograph, which is why I need some sort of probability based approach for this. My intuition says HMM might be applicable here, as there is an order in the projected colors, but I don't have enough experience with HMM to formulate this problem. Can anyone kindly point out a way I can achieve my result using HMM or otherwise?