A Novel Decision Fusion Method for Image Forgery Localization

To obtain more reliable localization performance, we propose a new image forgery localization scheme, which fuses the output of two diverse tools, based on PRNU and PatchMatch algorithm, respectively. First, tampering possibility maps are obtained from each of the two approaches, and then Dempster-Shafer theory of evidence is used as decision fusion strategy to obtain the final localization map. Comparison with single approach shows an improvement in tampering localization accuracy.