Recaptured Image Forensics Based on Quality Aware and Histogram Feature

摘要

The recaptured images forensics has drawn much attention in forensics community. The technology can provide some evidences for copyright protection and protect the face spoofing system to a certain degree. In this paper, we propose an algorithm to detect the images recaptured from LCD screen. On the one hand, the quality of the recaptured images would be affect in general. The generalized Gaussian distribution (GGD) and zero mode asymmetric generalized Gaussian distribution (AGGD) effectively capture the behavior of the coefficients of natural and distorted versions of them. So the parameters of GGD with zero mean and zero mode AGGD are estimated as the quality aware feature. On the other hand, the correlation of DCT coefficients between two adjacent positions would be changed. The histogram feature of difference matrix of DCT coefficients is used to measure it. The experimental results show that the proposed method obtains a outstanding detection accuracy.