Copy-Move Image Forensic Algorithm Based On SIFT

As a kind of multimedia information, digital images play an important role in the internet and people’s lives. With the rapid development of information technology, the powerful image editing software arise that makes image tampering much easier, and it becomes difficult to determine if a particular image is authentic or not by human eyes. Especially, when tampered images are used in some formal occasions, it would cause a series of confidence crisis. Therefore, the identification technology for digital images’authenticity and integrity becomes quite important, and has become one of the hotspots of current research directions. This paper mainly focuses on copy-move image forensics algorithms based on keypoints to improve their security and performance.Firstly, the research background and significance of digital image forensics are introduced, as well as the research situation at home and abroad. Then, we make a research and analysis on the keypoints-based forensic methods for copy-move attack detection, and point out their shortages, which is conducive to the proposal of new algorithms.For the existing copy-move forgery detection algorithms based on SIFT have a security threat, the forgery technique of hiding targets with flat regions is firstly proposed to attack the latest detection method based on SIFT. The experimental results show that unproper contrast threshold in extracting SIFT keypoints will break the credibility of the algorithm seriously. Then to effectively improve the ability against the attack of removing objects with flat regions, an improved scheme was given. Firstly, the contrast threshold was reduced to abtain enough SIFT keypoints in flat regions; then on the basis of a first clustering, much mismatch caused by the decreased contrast threshold was eliminated by RANSAC algorithm and a second clustering, Thus tampered regions are located accurately. While this kind of detection algorithm can not effectively distinguish the tampered image with copy-move attack and the honest image with two or more similar objects, so some false detection is caused.To reduce the false positives of copy-move attack detection algorithms based on keypoints for real pictures, an effective image passive forensic algorithm is designed in this paper. It is based on the characteristic that there is only a geometrical transformation in a plane betweeen the tamper areas, while there is three dimensional imaging disparity between the similar objects in true image. The improved algorithm is a SIFT-based forensic method, the positional relationships of matched SIFT points are used to judge whether there is only a geometric transform in a plane between these similar areas. if so, it’s false; otherwise, it’s true. Concretely, the center coordinates of all matching points in each suspected tamper areas are firstly computered, then calculating the vectors from all matching points to the center for every tampered block, and normalizing them. Lastly, a group of angles will be abtained from the above-mentioned vectors. By analyzing the fluctuations of these angles, we can distinguish the true image from the false image. Experimental results show that the proposed method can effectively decrease the error detection for true images.Lastly, a simulation system for the detection algorithms is designed, and the simulation results are analyzed and explained.