Joint Spatial and Discrete Cosine Transform Domain-Based Counter Forensics for Adaptive Contrast Enhancement

Abstract

Contrast enhancement (CE) is a common post-processing step in image forgery to create visually convincing tampered images. However, the artifacts embedded during this process can be captured to determine the presence of CE. To overcome these artifacts, we propose a novel counter-forensic technique using adaptive CE as an enhancement operation, whereas previous works only deal with global CE. We derive an optimization formulation, which enhances the attacked image using the L2 distance in both the spatial and DCT domains. The proposed algorithm suppresses the detectable artifacts, thereby reducing the CE detection performance. Furthermore, the formulation also preserves natural spatial statistics using Huber Markov random field. A major advantage of working jointly in both the domains is that the complementary information can be leveraged while suppressing the artifacts in both the domains. We evaluate the proposed method using various visual quality metrics and against the state-of-the-art CE detectors. In our experiments, we observe a reduction of more than 17% in accuracy for a false positive rate of 1% for deep learning as well as steganalysis-DCT feature-based detectors. We also show that the proposed model generates high visual quality images.