An improved reduced feature-based copy-move forgery detection technique

Abstract

With Widespread malpractice of copy-move image forgery, enforcing image forensics becomes imperative. In this approach, objects can be added into or removed from the same image to hide the truth with malicious intention. In order to address this issue, passive copy move forgery localization and detection has been playing a crucial rule in image forensics arena. The paper proposed a reduced feature-based algorithm which is robust as well as highly accurate in terms of detecting forged area. In this proposed scheme, stationary wavelet transform is employed on subject image to obtain low approximation band, and then significant features are extracted from it using block-based Discrete Cosine Transformation (DCT) and singular value decomposition (SVD) accordingly. Traditional block-based approach suffers from computational overhead especially for large images. Proposed scheme, extracts only three feature vectors, one DC component applying DCT on LL bands and two singular value components extracted employing SVD from the remaining AC components of each transformed block. These reduced features are further utilized for analysing and matching to detect identical regions in an image. In spite of having a smaller number of features, experimental results exhibit, this proposed scheme detects the forged area precisely as well as exhibits quality robustness against different post- processing attacks.