A Robust Approach for Crop Misalignment Estimation in Single and Double JPEG Compressed Images

Puglisi G.

Primo

;Battiato S.

2024-01-01

---

Abstract

In forensics investigation, information about the crop performed after a JPEG compression can be useful exploited to reconstruct the manipulation history of the analyzed images and localize forgeries. Statistics computed from AC histograms, obtained performing an additional JPEG compression with a set of constant quantization matrices, are exploited to design a robust and effective algorithm to retrieve information about the employed crop. Finally, the effectiveness of the proposed solution has been demonstrated through a series of tests conducted considering different patch sizes, quantization matrices, and comparisons with a state-of-the-art solution in single and double JPEG compression scenario.