Skip to Main Content
Copy forgery is the most commonly used operation of image forgery. This paper describes a new method based on LPP-SIFT(Locality Preserving Projection- Scale Invariant Feature Transform) features for image forgery detection. The algorithm first extracts SIFT keypoints of an image, and then combines LPP to obtain low-dimensional feature descriptors, the final stage is the keypoints matching. Each pair of matched keypoints of the image are marked with lines between them. If the image has undergone copy-forged operation, these lines will obviously concentrate upon two regions. Experiments demonstrate that the proposed approach is efficient for copy operation and other post-processing forgeries, such as rotation, scaling, and retouching.