Forgery and Double Compression Detection in Digital Images using Combined Features of Quantization Effects on DCT Coefficients

Recompression detection has an important role in JPEG image forensic and forgery detection. In this paper, a new algorithm is proposed for image forgery detection using combined features of quantization effect on DCT coefficients. In the proposed approach, we use distribution of most significant digit of DCT coefficients and features based on the period of DCT coefficients in three color channels. To select more proper features and reduce computational overhead of the algorithm, a new hierarchical approach is used for feature selection. The proposed feature selection algorithm can select both proper frequency region and more effective digit distribution for efficient forgery detection. To test the proposed algorithm and compare the results with those of other methods the forged and authentic images of CASIA database are used. The experimental results and a comparison of the results of the proposed algorithm with those of other method show the correct recognition rate of 99.5% as well as the enhancement of 7.6% with respect to existing approaches which demonstrate the efficiency of the proposed approach.