The Research On Detection Of Image Region-duplication Forgery

Currently, digital images have been widely used in people’s daily life and work. At thesame time, a lot of image editing and processing tools have developed so rapidly, that peoplecan tamper with the image content, to subvert the people "seeing is believing" the traditionalconcept. If the forgery and tampering of the digital image was used for scientific discovery,court exhibits, the official media, will cause great social and political impact. Therefore, fordigital image, tampering of evidence is of great significance. Digital image forensicstechnology can determine the authenticity and integrity of digital images based on providingcontents of the digital image itself, or statistical characteristics information.Targeting image region copy-move forgery, this paper proposed a detection algorithmbased on gray level co-occurrence matrix. Firstly, the detected image was divided intomultiple overlapping blocks with same size, represent the textural features of each block withthe statistics of its gray level co-occurrence matrix, and get the feature vector of the image.Secondly, we sorted the feature vector by dictionary sort, and located the tampered region byutilizing the displacement vectors of image blocks. Lastly, experimental results show thatalgorithm can detect copy-move tampering in terms of robustness against rotate operation andefficiency.In order to increase the difficulty of copy paste of the tamper detection, image area willusually be copied or further processing, such as scaling, rotating, brightness and adjustment,and fuzzy. Based on a LBP(local binary patterns) detection algorithm for arbitrary angle ofrotation and brightness adjusting copy paste is proposed. Firstly, all key points are extractedfrom the image by applying the SIFT(scale invariant feature transform) algorithm. Secondly,each key points is described by the rotation-invariant LBP patterns, which are computed fromthe image patch centered at the key point. Lastly, matching the key points by using thefeature vectors of Euclidean distance. The experimental results show that our algorithmperforms efficiency in terms of robustness against rotate operation and intensity adjustmentand higher matching accuracy.