I think I remember seeing a similar post about Chinese bootleg games for non-PC systems. The commonality was comparing pixel data to trace what the bootlegs had taken parts from. That raises a question: can we automate inferring the copying and modification of parts between older games?
In biology, there's a similar technique. Phylogenetic analysis compares sequences to reconstruct evolutionary relationships between samples. From what I understand, it seems possible to do something similar for the pixel data of game assets.
However, there are limits: you have to stay above a certain pixel size for results to have meaning. The lower your resolution, the more each pixel font looks the same. The same seems to be true for sprites representing in-game elements.