Re:How Perceptual Hashes Work
by jamii on Friday June 03, 2011 @01:47AM
(#2615305)
In the case of Hamming distance there is much more information to work with than in a bare metric. You can dump all the bitstrings into a suffix array, break the query into K pieces and search for an exact match for each piece. The resulting list of matches is usually small enough to just compute the Hamming distance against each element. I've used this on a corpus of 2m LaTeX strings withbgood results. See my above post for more details.
[ Parent
| on HN ]
There are no comments on this comment yet.