Detection of copy-move forgery in digital image using locality preserving projections

Abstract

Passive forensics is the technology for detecting the authenticity of the image content and the source of image without depending on any prior information. This paper proposes a method of passive forensic based on the locality preserving projection (LPP) mainly aimed at copy-move tampering in the same image. Firstly, put the detected image into several small blocks and carry on their coordinates. Then, on the premise that the similarity matrix guarantees the adjacent relation unchanged before and after the transformation, use the LPP algorithm to reduce the small block' s dimension. Finally, match the reduced dimension of the small blocks to complete the detection. Experiments show that this method can efficiently detect the copy-move forgery quickly and can accurately locate tamper with the position. Compare to the method of PCA proposed by Posucue, the proposed method effectively improves the detection precision with reducing the algorithm complexity and the testing time.