Image splicing detection based on Markov features in QDCT domain

Abstract

Image splicing is very common and fundamental in image tampering. Therefore, image splicing detection has attracted more and more attention recently in digital forensics. Gray images are used directly, or color images are converted to gray images before be processed in previous image splicing detection algorithms. However, most forgery images are color images. In order to make use of the color information in images, a classification algorithm is put forward which can use color images directly. In this paper, an algorithm based on Markov in quaternion discrete cosine transform (QDCT) domain is proposed for image splicing detection. First of all, color information is extracted from blocked images to construct quaternion in a whole manner, and the QDCT coefficients of quaternion blocked images can be obtained. Secondly, the expanded Markov features generated from the transition probability matrices in QDCT domain can not only capture the intra-block, but also the inter-block correlation between block QDCT coefficients. Finally, support vector machine (SVM) is exploited to classify the Markov feature vector. The experiment results demonstrate that the proposed algorithm not only make use of color information of images, but also can yield considerably better detection performance compared with the state-of-the-art splicing detection methods tested on the same dataset.