How Perceptual Hashes Work
hackerfactor.com
How Perceptual Hashes Work
1–10 of 77 posts
Re: How Perceptual Hashes Work
#2Thank you, I learned something.
Since TinEye pre-computes the hashes, do they use something like Redis to retrieve information? Redis seems perfect for a such quick results using the hash as the key and a URL or object of some kind as the value.
Re: How Perceptual Hashes Work
#3That was one of the best articles I've read in a while.
Re: How Perceptual Hashes Work
#4Wow, I had no idea it was so easy. Could this in fact be used to combat copyright infringement for the likes of The Oatmeal on large scale, given someone (Google) with the requisite computing power?
Re: How Perceptual Hashes Work
#5Thank you, I learned something. Since TinEye pre-computes the hashes, do they use something like Redis to retrieve information? Redis seems perfect for a such quick results using the hash as the key and a URL or object of some kind as the value.
It seems to me that what you want is some kind of spatial index - you might not get an exact match on the hash, but instead get one that's one or two bits away, and you'll want something better than linear search to find it.
Re: How Perceptual Hashes Work
#6Thank you, I learned something. Since TinEye pre-computes the hashes, do they use something like Redis to retrieve information? Redis seems perfect for a such quick results using the hash as the key and a URL or object of some kind as the value.
It seems to me that what you want is some kind of spatial index - you might not get an exact match on the hash, but instead get one that's one or two bits away, and you'll want something better than linear search to find it.
Doesn't scaling the images down (to 32x32 for the pHash approach) achieve essentially the same thing? Images that differ only slightly will likely scale down to the same thumbnail to begin with, and the resulting hash still bears some relationship to that thumbnail — so you should be able to look at similar hashes to find similar inputs.
Re: How Perceptual Hashes Work
#7"I'll use a picture of my next wife, Alyson Hannigan."
That had me doing a few google searches.
Re: How Perceptual Hashes Work
#8Earlier quoted context omitted.
It seems to me that what you want is some kind of spatial index - you might not get an exact match on the hash, but instead get one that's one or two bits away, and you'll want something better than linear search to find it.
Doesn't scaling the images down (to 32x32 for the pHash approach) achieve essentially the same thing? Images that differ only slightly will likely scale down to the same thumbnail to begin with, and the resulting hash still bears some relationship to that thumbnail — so you should be able to look at similar hashes to find similar inputs.
Often you get things like images that have had text added to them (e.g. I often use tineye to find the original source image that someone has added a caption to), which means that the thumbnails will differ to some degree.
Re: How Perceptual Hashes Work
#9See also SIFT
Re: How Perceptual Hashes Work
#10Earlier quoted context omitted.
It seems to me that what you want is some kind of spatial index - you might not get an exact match on the hash, but instead get one that's one or two bits away, and you'll want something better than linear search to find it.
Doesn't scaling the images down (to 32x32 for the pHash approach) achieve essentially the same thing? Images that differ only slightly will likely scale down to the same thumbnail to begin with, and the resulting hash still bears some relationship to that thumbnail — so you should be able to look at similar hashes to find similar inputs.
That's exactly what I'm saying - a simple index will only find exact hash matches, but what you want is images with "similar" hashes. If you define "similar" as a low Hamming distance - i.e. a small number of bit differences - and you want to find these image hashes with something better than a linear search / exhaustive combinatorial bit-twiddling, then you'll need to be smarter about your index.