Hany Farid coined the term "JPEG Ghosts" for distinguishing single- and double-JPEG compressed image regions. However, exploiting this observation required a tedious manual investigation of a large number of grayscale-images.

In our work "Automated Forgery Detection Through Classification of JPEG Ghosts", we propose a pattern recognition scheme to achieve fully automated, highly accurate detection of double-compressed JPEG blocks.

The paper is available here.

The code can be downloaded here.

If you use the code, please cite our paper
Fabian Zach, Christian Riess and Elli Angelopoulou: "Automated Image Forgery Detection through Classification of JPEG Ghosts", Pattern Recognition (Joint 34th DAGM and 36th OAGM Symposium), Graz, Austria, August 2012, pp. 185-194.

For our evaluation, we used the Uncompressed Colour Image Database (UCID) by Schaefer and Stich.