A Statistical Property Based Expanding Block Algorithm For Copy-move Forgery Of Digital Image

With the rapid development of multimedia processing, network technique and digital communication technology, digital image has been acting as an important carrier of information spreading and interaction. Meanwhile, the popularization of some software tools such as Photoshop makes it more and more convenient to edit and modify digital images. For one thing, this makes images presented around us with better superior visual effect, hence makes images better meet people's various demand. For another, these tools may also turn to be tools for somebody with ulterior motives. Based on this, some kind of method is needed for detecting digital image forgery. Therefore, the digital image forensics technology emerged.Within the various tampering of digital image, the copy-move forgery appears to be one of the most important and popular means of tampering of digital image. This paper focuses on this type of digital image forgery and makes deep research on this topic. The existing methods such as exhaustive search, autocorrelation, method based on DCT, PCA and SIFT are studied in this project.After studying this existing algorithms and analyzing the shortcomings and weakness of them, a new expanding blocking algorithm is proposed based on statistical property in this paper. This algorithm is brought in on the basis of chi square distribution theory and hypothesis testing theory in the theory of probability and statistics. Comparing with the existing algorithms, the proposed method performs as well as them, in some respects may better than them. Moreover, this algorithm does well when the forged region has been manufactured under JPEG compression, with the effect of Gaussian blurring or slightly made lighter or darker. Especially when the forged region is made slightly lighter or darker, this algorithm can perform with a unique advantage.