Let us consider the problemm of separating two imbricated peaks (A and B) in a spectrum without any a priori knowledge of the shape pprofile of the two given peaks.
I think that Bayesian and max entropy methods are appropriate ; but I have no idea the way to carry them out. I think one has to consider, for a given abscissa that (supposing for simplicity the ordinates are integer ranging 1 --> 256)
Probability (A=1) and Probability (B=255) Probability (A=2) and Probability (B=254) and so on ...
And let us not care to the fact that computations are very heavy.