A copy-move detection algorithm based on binary gradient contours

Copy-move is one of the most obvious ways of deliberate distortion of digital images in order to conceal the information contained in them. The process of duplicate embedding consists in copying an image fragment and pasting it within the same image. Prior to pasting, the fragment can be distorted using transformations such as contrast enhancement, noise adding, scaling, rotation, and combinations thereof. Existing approaches to copy-move forgery detection are based on calculating feature vectors for overlapping blocks of an image and then using these vectors to find the closest regions in Euclidean space. In this paper, we propose features based on binary gradient contours, which are resistant to contrast enhancement, additive noise and JPEG compression. We also present results of conducted experiments for demonstrating the proposed algorithm effectiveness for a range of distortion parameters. The research also involves comparing features based on binary gradient contours with features based on various forms of local binary patterns.