A robust technique for copy-move forgery detection from small and extremely smooth tampered regions based on the DHE-SURF features and mDBSCAN clustering

• Published in: Australian journal of forensic sciences Vol. 53; no. 4; pp. 459 - 482
• Main Authors: Bilal, Muhammad, Habib, Hafiz Adnan, Mehmood, Zahid, Yousaf, Rehan Mehmood, Saba, Tanzila, Rehman, Amjad
• Format: Journal Article
• Language: English
• Published: Clovelly Taylor & Francis 04.07.2021; Copyright Agency Limited (Distributor)
• Subjects: Clustering; CMF detection method; COSTS; DBSCAN clustering; Digital images; Equalization; Euclidean geometry; Feature extraction; FORENSIC SCIENCE; Forensic sciences; Forgery; Histograms; Image contrast; Image manipulation; JPEG (Image coding standard); Robustness; SURF features
• ISSN: 0045-0618; 1834-562X
• DOI: 10.1080/00450618.2020.1715479

The keypoint-based copy-move forgery (CMF) detection is one of the most widely used CMF detection methods; however the keypoint-based CMF detection method cannot effectively and efficiently detect the small and extremely smooth tampered regions in the input image. A CMF detection method is proposed to tackle the above mention problem. In the proposed CMF detection method, the contrast of the input image is adjusted using the dynamic histogram equalization (DHE) method. A speeded-up robust feature (SURF) descriptor is used to extract features from the tampered image and matched using Euclidean distance. The novel modified density-based spatial clustering of application with noise (mDBSCAN) clustering technique is applied to the matched features to generate the binary mask followed by the detection of CMF regions. Three standard datasets, MICC-F220, MICC-F2000, and CoMoFoD, are used to evaluate the proposed CMF detection method performance. The experimental results indicate that the proposed CMF detection method outshines the state-of-the-art CMF detection method in terms of precision (P) and recall (R).