0
$\begingroup$

Could anyone enlighten me on what are the current best ways to detect if two images are the same (i.e. a copy or edited copy of each other)? What feature detection method do they use (SIFT, DWT, pixel intensity, etc.)?

I've read a lot of algorithms that focus on detecting similar objects and so on, but I am interested in reading on some for detecting whether two images are copies of each other. I've done some googling and most algorithms/papers I've found are pre-2014 or are about detecting similar images, not exact copies.

$\endgroup$
  • $\begingroup$ If you allow edits, how would you classify an image were multiple pastes from different images? $\endgroup$ – Stanley Pawlukiewicz Jun 2 '18 at 1:27
  • $\begingroup$ Maybe send someone to do CRC check? $\endgroup$ – mathreadler Jun 2 '18 at 7:56
  • $\begingroup$ Maybe send someone to do CRC check? $\endgroup$ – mathreadler Jun 2 '18 at 7:57
  • $\begingroup$ @StanleyPawlukiewicz by edits I mean much simpler edits, i.e. brightness change, change to black n white, rotation, etc. $\endgroup$ – Daniel Jun 3 '18 at 4:11
0
$\begingroup$

This is not a simple problem. You need to define these 'simple edits' and check for all of those.

For example if you want to find out brightness changes, you cannot effectively use template matching, you would have to go with feature matching (contour, keypoint etc.).

For a general solution, you need to extract a 'signature' from an image which for instance can be done with a combination of histogram and feature points (edge, corner, blob, Shi-Tomasi, LoG etc.)

$\endgroup$
0
$\begingroup$

As noted in the previous answer, you need to extract a signature. A simple approach is to compute a "perceptual hash" (pHash) of the image. From pHash.org:

What is a perceptual hash? A perceptual hash is a fingerprint of a multimedia file derived from various features from its content. Unlike cryptographic hash functions which rely on the avalanche effect of small changes in input leading to drastic changes in the output, perceptual hashes are "close" to one another if the features are similar.

$\endgroup$

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.