I don't think anyone has a definition of "nearest" that could accomplish that in general. Comparing pixel data is easy, but comparing the subjects portrayed and how is much harder to pin down.