I often have many images, but I only need one of each. I've used DoubleKiller, but the problem with it is that it uses the CRC32 checksum of the files to tell if they are duplicates, which is problematic because it will only find exact, byte-by-byte duplicates, but not two images that are the same but have different metadata, or different compression ratios, or different file types, etc. So it's not very effective for me. So can anyone recommend a free (preferably open source) Linux program that can do that (I can handle using a Windows program if it at least runs through Wine) program that is able to detect similar images based on visual similarities instead of file contents? Or even better, does GIMP or another pre-installed program have that capability?

I'm thinking of something like this, although I don't really trust that program (I don't like CNET in general), plus it's not for Linux.