A two-stage detection method of copy-move forgery based on parallel feature fusion

Abstract The copy-move forgery refers to the copying and pasting of a region original image into target same image, which represents typical tampering method with characteristics easy high-quality tampering. existing single feature-based methods detection have certain shortcomings, such as high false alarm rate, low robustness, accuracy. To address these this paper proposes an improved two-stage based on parallel feature fusion adaptive threshold generation algorithm. Firstly, SLIC super-pixels segmentation algorithm is used for preprocessing, similar extraction without employed obtain suspected regions similarity. Secondly, obtained SIFT HU features express local regions. Then, corresponding value generated histogram oriented gradient (HOG) describe texture regions, acts criterion judge whether has been forged or not. experimental results show that proposed outperforms methods, achieving accuracy 99.01% 98.5% MICC-F220 MICC-F2000 datasets respectively. In addition, stronger robustness performance COMOFOD dataset than comparison methods.