Description

A novel copy-move forgery detection scheme that can accurately localize duplicated regions with a reasonable computational cost. To this end, a new interest point detector is proposed utilizing the advantages of both block-based and traditional keypoint-based methods.The detected keypoints adaptively cover the entire image, even low contrast regions, based on a uniqueness metric. Moreover, a new filtering algorithm is employed, which can effectively prune the falsely matched regions. Considering the new interest point detector, an iterative improvement strategy is proposed. The whole procedure is iterated along with adjusting the keypoints density based on the achieved information. The exper-imental results demonstrate that the proposed scheme outper-forms the state-of-the-art methods using two public benchmarkdatabases.