If a jpeg is (tampered with and) saved again as jpeg, you get 'Double Quantization' effects.
I would strongly suggest Hany Farid's homepage (given by Joachim. He is one of the best resources in this area. I would also suggest
Passive-blind Image Forensics (2006) by Tian-tsong Ng , Shih-fu Chang , Ching-yung Lin , Qibin Sun at http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.83.7689
One additional thing. If this is to testify, be careful in relying in just one technique. There is already a precedent in US jurisdictions that the testimony in image forensics needs to have several tests to confirm the findings. Again this may rely on the level of tampering you may need to rely on a different set of techniques. For example if there is metadata tampering it is not just a matter of determining an inconsistency but the overall temporal analysis made that gives you the answer whether tampering occurred.
In one case that I had we had two levels of validation and based on the case we were prepared to use both of them. The case was a video showing a person with specific body marks. I did the initial artifact test validation and concluded that based on the compression it was not an artifact issue. Nonetheless, I had another colleague ready to do scene analysis to determine the height of the person should my test not be enough.
Hope this helps
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing