Robust copy-move forgery detection method using pyramid model and Zernike moments

Abstract

Copy-move forgery detection (CMFD) is probably one of the most active research subtopics within the blind image forensics field. Among existing algorithms, block-based methods achieve better detection result, but suffer from not being robust to the scaling of the copied region, which is limited to a small scaling range between 91 and 109%. To overcome this issue, a novel copy-move forgery detection method based on the pyramid model and Zernike moments is proposed in this work. First, a non-uniform sampling approach is used to build the pyramid model to improve the speed of detection, where each level of the model is a rescaled image with different scale factors. Then, the Zernike moments method is used to extract image rotation invariant feature. Furthermore, to make the proposed model be robust to the rotation and the combination of the rotation and scaling of the copied region, the random sample consensus (RANSAC) algorithm is adopted to look for the same affine transformation parameters. Compared with the existing methods, experimental results show that the proposed method has a good effect on the arbitrary rotation angle, the scaling range between 50 and 200%, and the combination of both.