An Evaluation of Popular Copy-Move Forgery Detection Approaches

摘要: A copy-move forgery is created by copying and pasting content within the same image, potentially postprocessing it. In recent years, detection of forgeries has become one most actively researched topics in blind image forensics. considerable number different algorithms have been proposed focusing on types postprocessed copies. this paper, we aim to answer which processing steps (e.g., matching, filtering, outlier detection, affine transformation estimation) perform best various scenarios. The focus our analysis evaluate performance previously feature sets. We achieve casting existing a common pipeline. examined 15 prominent analyzed per-image basis per-pixel basis. challenging real-world dataset, software framework for systematic manipulation. Experiments show, that keypoint-based features Sift Surf, as well block-based DCT, DWT, KPCA, PCA, Zernike very well. These sets exhibit robustness against noise sources downsampling, while reliably identifying copied regions.