image Step I : Keypoint extraction Step II : Keypoint matching Step III : T-Linkage clustering Step IV : Pyramidal decomposition Step V : Multi-scale analysis Step

Yi Fan,Yue-Sheng Zhu,Zhen Liu

2016-01-01

Abstract:The detection of copy-move forgery is an important module in the image authentication. It aims to detect whether an image contains copy-move tampering and where the tampered regions are. To improve the detection accuracy, in this paper we propose to apply local feature matching, followed by T-Linkage and multi-scale analysis by employing the T-Linkage algorithm for robust clustering in the space of geometric transformation. We further capture coarse-to-fine information by constructing a multi-scale image representation and locating duplicated regions for each scale. We make a final decision by considering all the localization maps with a voting strategy. Experimental results show that the proposed method achieves desirable detection and localization performance on different copy-move attacks.