Is there an image search engine that can.....
 

Take an image you give it, and find the exact same image on other sites?

I think for that to work, the search engine in question would have to actually get down into the binary code of the image or something. Not really sure how that could work.

The only way I see something like that working would be to find the same image IF it were named the same. i.e. Foo.jpg = Foo.jpg or Foo.gif or Foo.bmp etc.

But then you are still going to end up with a bunch of pictures that really have nothing to do with one another. I mean, how many pics named "assfuck.jpg" are on the web you think?

Finding the exact same image would probably be possible, yes.

That would take a hash search, I don't think there's one out there for public use, it's something similar to what the government uses to find CP pics.

Would that only find the image if it had remained intact as far as size and format goes though? Like say I took an image that was 640x480 and resized it to half that and saved it in gif format, wouldn't that change the pictures actual internal code? How does that work?

If the pics are of same size and everything, then you can compare their hash values (I don't know of any engine that currently does that, but it might be out there, dunno). However, it will fail to find the same image if it's been resized, compressed, or modified in any other way.

This is completely possible.

Say the search engine crawler saved the images it came across as a 0-9 number hash, where the numbers represent the approximate color of the pixel in question (it would first have to greyscale + increase the contrast of the image because each pixel only has 10 variations for its representation). Then the search query would generate another hash of the inputted image, and use the levenshtein principle (distance between two strings) to find its closest matches in the database.