pHash (perceptual hash) for duplicates & similar

[Goulu]

Hello,

I suggest indexing all images in the database by a "perceptual hash" . See here : http://www.hackerfactor.com/blog/?/arch ... ke-It.html it's extremely simple and fast. There is a library here to do it : https://www.phash.org/

I use it for a professional project and it's incredible : with a 32 bits key, you can instantaneously (well O(log N) ...) find duplicates in hundreds of thousands of images. And very quickly find similar images.

Since XnView already builds a database, I think this feature would make it the fastest duplicates and similar image finder on Earth.

[xnview]

thanks, i'll check it

[m.Th.]

AHEM!

http://newsgroup.xnview.com/viewtopic.php?f=60&t=28396

[jadO]

thanks, i'll check it

Will that be built into XnViewMP?

[xnview]

yes usefull, but to create it, i need to load image

[jadO]

yes usefull, but to create it, i need to load image

Is that a problem? From my understanding when a creating thumbnails, the image must be loaded too. So why not combining thumbnailing with hashing?

[xnview]

yes usefull, but to create it, i need to load image

Is that a problem? From my understanding when a creating thumbnails, the image must be loaded too. So why not combining thumbnailing with hashing?

No, image is not always loaded, when embedded thumbnail is available or small preview