A New Approach Merging Markov and DCT Features for Image Splicing Detection

Splicing detection is of fundamental importance in digital image forensics. Recent image forensic research has resulted in a number of tampering detection techniques utilizing statistical features. Fusion of multiple features provides promises for improving detection performance. In this paper, we propose a new splicing detection approach based on the features utilized for steganalysis. We merge Markov process based features and discrete cosine transform (DCT) features for splicing detection. The proposed approach can achieve an accuracy of 91.5% with a 109-dimensional feature vector. Experimental results demonstrate its superior performance over the prior arts.