3 ms·
What I find strange is the use of a hash of the image. There are other methods to detect whether images are visually the same while they have a different resolu
by mosselman 11y ago
What I find strange is the use of a hash of the image. There are other methods to detect whether images are visually the same while they have a different resolution, etc.
From what I remember is that you'd resize all images to a fingerprint of yxy (lets say 100x100 pixels), apply some filtering in order to normalize lighting and then you'd XOR 2 fingerprints in order to see how much they are similar.
- lmm 11y agoThat's more complicated, more likely to false-positive, and I doubt it would provide much more value. Very few people are resizing images and uploading a slightly different version; the overwhelming majority of the time, duplicate images are people uploading the exact same file, byte-for-byte.