Image Splicing Detection Based on Improved Markov Model

Digital image splicing detection is a new and important subject in image forensics. Research shows that Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT) based Markov features are effective for image splicing detection. However, the state selection in the traditional Markov model was simply rounding the parameters and taking threshold value, which has not exploited the parameter distribute information. In this paper, a novel Markov state selection method is proposed. The approach matches states with parameters evenly according to fixed ratio calculated by pre-set state numbers. Experiments show that the improved Markov model achieves higher recognition accuracy rate compared with the traditional Markov model with the same feature dimension.