Detection Method Based On Surf's Copy-move Tampering

The image tampering is becoming more and more easily due to powerful image processing software. There are many kinds of method of digital image tampering, one is Copy-move modification in which a part of the image is copied and pasted somewhere else in the image. It is easy to realize the Copy-move tampering as well as good performance. Distorted image may cause serious adverse effects to economic and social life.Firstly, we simulated and analysis the performance of existing typical tamper detection algorithm of Copy-move attack. Then we gave a suitable Copy-move tamper model of multi-zone and geometric transformation; in view of the problem that the existing Copy-move tamper detection algorithm have poor robust on geometric transformation and heavy computation, a new tamper detection method was proposed. The new method using SURF(Speed Up Robust Feature) algorithm to get the interesting points from the distorted image and generate the descriptors of interest points. We take Euclidean distance between the point descriptors as the basis of the interesting points matching. According to the characteristics of Copy-move tampered, Clustering algorithm was used to purify the matching interesting point. All the matching points were linked by lines located between the two regions obviously in a Copy-move tampered image. The minimum circumscribed rectangle of the matching point supposed to be the tampered area.300 piece of images from CASIA set was experiment to test the performance of algorithm, it comes out that the correct detection rate is 97%. The experiment results show that the algorithm can detect multi tamped zone, effective resistance to scaling, rotation geometric transformation. It is also undergone some attack, such as loss JPEG compression Gaussian Blur filter, Gaussian White noise pollution and their hybrid attack. The result also demonstrated that the efficiency of the proposed algorithm is good.