A few years ago, the findimagedupes program by Rob Kudla found its way into Debian. I noticed it, played with it a bit, and filed some bug reports / suggestions for improvement. Then I forgot about it.
Some time later, I was contacted by the Debian maintainer, asking if I would submit my improvements. It seemed simpler to rewrite the code than to patch it. So, after many delays and interruptions, that's what I did.
My version of findimagedupes has now replaced Rob Kudla's code in Debian.
Documentation is built-in:
Alternatively, you may view a preformatted manpage produced by pod2html.
The code is also available from GitHub.
You'll need Perl, GraphicsMagick and a few CPAN modules. See the beginning of the source for details.