Yeah, it does demonstrate that you can use compression to measure similarity of two images. But it also demonstrates that it's a pretty poor similarity measure

Yeah, it does demonstrate that you can use compression to measure similarity of two images. But it also demonstrates that it's a pretty poor similarity measure. Something as simple as counting % of matches between the black and white pixels performs much better.