In our application, our users submit a lot of images which are actually just resized images of existing images, and need to catch all such duplicate images. I have done a comparison for the accuracy of pHash DCT vs MH for resized images and would like opinion on the same.
but my problem doesn't get solve
Checking for Different Images (i.e Non-Duplicates)
I first took a small set of images (name "Green Boards") which are all different images (but are very similar looking). I wanted these images to be marked as non-duplicates by pHash.
I even then resized all above images into a common size of 200 as well as common size of 160
On checking all combinations for these images, I noticed that I are seeing DCT values as low as "2" and MH values as low as "0.057292".
Is there a way to make it work only with either DCT or only with MH? Or is there another Hashing type I should be using?