Recaptured image forensics based on normalized local ternary count histograms of residual maps

Image display and acquisition technologies have greatly developed over the past few years, which make it possible to recapture high-quality images from high-fidelity LCD (liquid crystal display) screens. These recaptured images pose serious threats on current image forensics technologies and various intelligent recognition systems, which makes the detection of such images significantly important. In this paper, we propose a recaptured image detection method based on normalized LTC (local ternary count) histograms of residual maps. Specifically, given an image, the residual maps of itself and its downsampled version are calculated in gray space via pixel-wise adaptive Wiener filtering. Then, the normalized LTC histograms of all the residual maps are extracted and concatenated as the final feature for training and testing. Single database experiments demonstrate that our proposed method not only performs very close to the state-of-the-art methods on relative low-quality NTU-ROSE and BJTU-IIS databases, but also improves the performance on the most difficult-to-detect ICL-COMMSP database obviously, which verifies the effectiveness of the proposed algorithm. Besides, mixed database experiments verify the superiority on generalization ability of our proposed method. Moreover, it has a lower computational complexity.