A probabilistic framework for copy-move forgery detection based on Markov Random Field

Copy-move forgery is one of the most common kind of image tampering where some part of an image is copied, may be with minor modifications, pasted to another area of the same image. With the growing usage of images in todays life, image authenticity has become a vital issue and consequently many image forgery detection techniques have been presented. In this paper, for the first time, we propose to treat copy-move forgery detection as labeling problem in a Markov Random Field. To gain a proper balance between precision and speed, an over segmentation is performed as a preprocessing step to obtain superpixels which are then regarded as nodes of the markov network. Intelligent selection of unary and binary potentials let the maximum a posteriori labeling to be a precise map of the forged regions. Qualitative and quantitative comparison with the state-of-the-art methods using public benchmarks demonstrate that the proposed method can improve precision while keeping the processing demands low.