Robust copy-move forgery detection based on multi-granularity Superpixels matching

摘要: In this paper, we propose a new multi-granularity superpixels matching based algorithm for the accurate detection and localization of copy-move forgeries, which integrated advantages keypoint-based block-based forgery approaches. Firstly, divide original tempted image into non-overlapping irregular coarse-granularity superpixels, stable keypoints are extracted from each superpixel. Secondly, superpixel features, is quaternion exponent moments magnitudes, superpixel, find (suspected region pairs) rapidly using Exact Euclidean Locality Sensitive Hashing (E2LSH). Thirdly, suspected pairs further segmented fine-granularity within replaced with superpixels. Finally, neighboring merged, obtain detected regions through morphological operation. Compared state-of-the-art approaches, extensive experimental results, conducted on public databases available online, demonstrate good performance our proposed even under variety challenging conditions.