Identifying Image Splicing Based on Local Statistical Features in DCT and DWT Domain

In this paper, an effective image splicing detection algorithm based on the local statistical features in DCT and DWT domain is proposed. The local ternary pattern (LTP) operator is introduced to characterize the statistical changes of DCT and DWT coefficients caused by image splicing. The LTP histograms are generated from the magnitude components of the block DCT coefficients with varying block sizes and the DWT coefficients in three detail subbands, respectively. All these LTP histograms are concatenated together to form the discriminative feature set for splicing detection. The effectiveness of the proposed detector is evaluated on the Columbia image splicing detection evaluation dataset. Simulation results have shown that the proposed method can perform better than several state-of-the-art methods investigated.