1
$\begingroup$

I have an image depicting a results of gel electrophoresis (it does not really matter - it just means: no bodies, faces, predictible objects) and I have a strong suspicion that a part of the image background was copied (probably Photoshop Clone Stamp) to cover an unwanted peak. The background is really just noise, but it should be almost perfect copy. I am not entirely sure how big is the copied area so just straightforward cross-correlation with an image part won't work.

How do I prove that? I prefer Matlab for such tasks.

$\endgroup$
2
  • $\begingroup$ clone stamp means no rotation or scaling, so autocorrelation of the image should show a peak when the two copies line up. probably high-pass everything first to emphasize just the background noise and edges lining up with each other, and not unrelated bars? $\endgroup$ – endolith Jun 3 '16 at 21:00
  • $\begingroup$ Votes and best answer validation are required for this question $\endgroup$ – Laurent Duval Jul 28 '19 at 12:07
2
$\begingroup$

Maybe this is an appropriate paper for you: Exposing Digital Forgeries in Scientific Images, by Hany Farid (freely available, e.g. at http://www.ists.dartmouth.edu/library/199.pdf).

$\endgroup$
0
$\begingroup$

You can compute different moments and statistics on sliding windows, build a multivariate image from them, and detect multivariate strong variations, perhaps with multidimensional operators, possibly segmentation tools like Geodesic Active Regions (level sets).

Hard to know without an idea of the image and the noise

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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