Consider multiple images that share a common feature. More specifically, they are the sum of multiple features, and I know that one of those is present in all. What is the best method to extract this common ground? My naive approach would be to take the minimum over all images for every pixel, but I wonder whether it can be done better.
I am mostly working in Fourier space, but this being a local problem, I expect it to be solved better in real space? A solution in Fourier space would be ideal, though.
For an illustration, see the following images:
What if Goofy was translated and/or rotated?
