Fast copy-move forgery detection using local bidirectional coherency error refinement

In this paper, we present an algorithm that can accurately and robustly detect regions of copy-move forgery. We firstly adapt and enhance a coherency sensitive hashing method to establish the feature correspondences in an image. Then, a local bidirectional coherency error is proposed to refine the feature correspondences via iteration over the enhanced coherency sensitive search. When the variation in the local bidirectional coherency error of the host image is not larger than a specified threshold, the iterative process stops, indicating that the feature correspondences are stable. In the end, from the stable feature correspondences, the copy-move forgery regions are easily detected using the local bidirectional coherency error of each feature. The experimental results show the proposed detection method achieves real-time or near real-time effectiveness; at the same time, it can achieve very good detection results compared with the state-of-the-art copy-move forgery detection algorithms, even under various challenging conditions. (C) 2018 Elsevier Ltd. All rights reserved.