A Copy-Move Forgery Detection Scheme with Improved Clone Region Estimation

To estimate a clone region in a digital image is an important task of copy-move forgery (CMF) detection, which is one of the key techniques of digital image forensics. Most existing keypoint-based CMF detection schemes are weak in distinguishing the clone regions and the similar regions. However, some misjudgments may be resulted in some similar regions, which are not the clone regions. To solve this problem, a novel copy-move forgery detection scheme is proposed. To avoid misjudgments, rough clone regions are estimated first, which can obviously reduce the false positive rate. And then the customized threshold is created in accordance with characteristics of each image, which can reduce the false negative rate as far as possible. Experimental results show that our proposed scheme obviously restrains the interference by the similar regions and has an outstanding performance on estimating clone regions.