Neat... I can see a use case for this in comparing pre- and post-final HTML documents (though I don't relish the headaches involved in actually acquiring both relevant sets of images at the same time out of our CMS).
Presumably it gets deeply murky when dealing with compressed jpegs?