rhsimplex / image-match

🎇 Quickly search over billions of images
2.94k stars 405 forks source link

Why "An image signature for any kind of image, Wong et al" instead of pHash? #20

Closed mladdict closed 8 years ago

mladdict commented 8 years ago

Thanks for the awesome package. Is there a rationale for choosing to implement the digital signature from "An image signature for any kind of image, Wong et al" versus pHash?

rhsimplex commented 8 years ago

Glad you find it helpful. To answer your question: part of it was ignorance of the field, and part of it was licensing.

To the first part, we looked into pHash, but I didn't understand how to make it scale to 1B+ images at the time. The Wong paper addresses that. Of course, now I see that the database method of storing the signatures could work for any hashing library, including pHash.

As for licensing, we weren't sure how we were going to distribute image-match and the pHash library uses GPL 3, which was a little too restrictive.

I would be really nice if we had plugins for other hashing libraries, certainly.