Some ads, like these two:

http://pagead2.googlesyndication.com/pagead/imgad?id=CICAgKDj2__aTRABGAEyCG3qbJztYSV0

https://tpc.googlesyndication.com/pagead/imgad?id=CICAgKDj2__aTRABGAEyCG3qbJztYSV0

are actually the same image served from different sources (with different URLs).

We want them stacked (as duplicates) to prevent cases like this:

We might try to compare them first by dimensions, then by file size, possibly even by their ID (as seen in this example, though the subdomain is different, the ID is identical), and finally, if these are not reliable enough, compare them bitwise to make sure they're one and the same.