Copy-move forgery detection using adaptive keypoint filtering and iterative region merging

Abstract

Copy-move forgery detection can generally be divided into two categories: block-based or keypoint-based methods. However, the existing block-based methods are usually lack of efficiency and the keypoint-based methods have not good detection performance. In this paper, a novel method using the adaptive keypoint filtering and iterative region merging is proposed for copy-move forgery detection. First, a feature extraction algorithm is presented to obtain the candidate keypoint pairs. Subsequently, adaptive keypoint filtering involving adaptive nearest neighbor pair filtering and outlier filtering is proposed to remove the outliers and obtain the inlier (authentic keypoint) pairs. The iterative region merging involving adaptive region iteration and region merging is proposed to iteratively generate more neighboring keypoint pairs and then merge the image segmentations (superpixels) to implement the copy-move region matting. Compared with other state-of-the-art methods, a series of experiments show that the proposed method can overcome defects and achieve better efficiency while keeping the high detection precision in copy-move forgery detection even under conditions that include various post-processing distortions.