Image splicing localization based on CFA-artifacts analysis

Image splicing is a widespread image forgery technique in which fragments from another image are pasted into the image under forgery. In this paper, a method of image splicing localization based on the analysis of CFA-artifacts that appear in the image during the capturing process is described. A feature characterizing the presence/absence of CFA artifacts for each image block is measured. The obtained values of the feature define the probability of each block to be embedded. Analysis of the accuracy of the splicing localization method and its robustness against different types of tampering, such as additive Gaussian noise, JPEG compression, and linear enhancement are presented in the experimental part of the paper. The results show that the suggested method reveals the embedded regions of different shape, size, and nature in images. The method is found to be stable to the additive Gaussian noise and linear enhancement, but not stable to JPEG compression. The advantage of the method is the ability to localize the spliced-in regions as small as a $2\times 2$ block.