Copy-move tampering detection using keypoint based hybrid feature extraction and improved transformation model

Digitally tampered images or fake images when propagated across the Web and social media, have the power to mislead, emotionally distress and influence public attitudes and behavior. Copy-Move tampering is one of the most commonly used attacks for damaging the semantics of an image. Keypoint-based methodologies are one of the effective ways of identifying a copy-move attack on an image. Existing key-point based methodologies fails to obtain a sufficient number of points on the small-smooth tampered region. Thus, for obtaining a good number of features this paper presents Hybrid Feature Detection (HFD) methodology employing Speeded-Up Robust Features (SURF) and Scale-Invariant Feature Transform (SIFT) descriptor. Further, hierarchical clustering optimization and an improved mismatch elimination model is presented for detecting and demarking tampered segments. Experiments are conducted on standard datasets such as Dataset (D0, D1-2 and D3), GRIP, MICC-F600 and MICC-F8Multi. Better results are achieved from the proposed HFD methodology when compared with existing state-of-art tampering detection methodologies in terms of Recall, F1-score, Precision and False Positive Rate (FPR).