Double JPEG detection using high order statistic features

Abstract

In this paper, we propose a reliable method to detect the presence of double JPEG compression. This method is based on content analysis and high order statistic features. Firstly, the image is divided into texture region and smooth region according to its content. Then, the difference arrays of the texture regions are calculated in order to enhance double JPEG compression artifacts. For better modeling the difference arrays, we improve the traditional first-order Markov transition probability algorithm and use second-order and third-order Markov transition probability matrices. In addition, a thresholding technique is used to reduce the size of the transition probability matrices. By using this method, we obtain a high order feature of each JPEG image for detection. And these features are fed into classifiers. Experiments have shown that our proposed method is effective.