A LBP-Based Method for Detecting Copy-Move Forgery with Rotation

Copy-move is the most common tampering manipulations, which copies one part of the image and pastes into another part in the same image. Most existing techniques for detecting tampering are sensitive to rotation and reflection. This paper proposed an approach to detect Copy-Move forgery with rotation. Firstly the suspicious image is divided into overlapping blocks, and then LBP operator are used to produce a descriptor invariant to the rotation for similar blocks matching. It is effective to solve the mismatch problem caused by the geometric changes in duplicated regions. In order to make the algorithm more effective, some parameters are proposed to remove the wrong matching blocks. Experiment results show that the proposed method is not only robust to rotation, but also to blurring or noise adding.