Interesting! I am always learning, and you are an excellent mentor.
Are you aware of this paper?
http://phash.org/docs/pubs/thesis_zauner.pdf
It's what, you guessed it, Phash.org tools is based on.
J
Perceptual Hash
-
fmw42
- Posts: 25562
- Joined: 2007-07-02T17:14:51-07:00
- Authentication code: 1152
- Location: Sunnyvale, California, USA
Re: Perceptual Hash
Yes, that was the first paper that Magick sent to me when I got started. But I could not figure out all of what they did and IM does not have a DCT nor a Radon transform (except within -deskew). Some of those are terribly simple and not very effective. For example the block hash was what Magick first thought to implement, but it is not very effective. And they do not explain how they get the hash for the Marr-Hildreth method, though IM has the functions to compute it via -morphology convolve DoG or LoG.
Have you compared any of those methods with my phash over a large range of images (or even just my test set)?