Font Size: a A A

Research On Detection Of Digital Image Region Duplication Forgery By Passive Authentication Methods

Posted on:2015-01-08Degree:MasterType:Thesis
Country:ChinaCandidate:T WangFull Text:PDF
GTID:2268330428964736Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Recently, with the rapid development of the multimedia technology and the widely employment of high-definition digital camera and image processing software, more and more people can obtain, edit, modify the images simply and conveniently. In this case, the images we get may no longer believable. Meanwhile, there are many methods to forge the image contents. The most popular method is called region duplication forgery which is very easy for people to forge the image contents. Thus this forgery scheme has been widely used by the unauthorized people. In this background, this thesis tries to develop a set of solutions to verify the region duplication forgery based on this typical digital image forgery-region duplication forgery authentication.We firstly review some typical image forgery methods. Then the digital image passive authentication technology is discussed in detail, especially the region duplication forgery authentication technology. By reintroducing the definition of the models of region duplication forgery, we give the disadvantages and advantages of existing algorithms. Consequently, we propose two specific methods to solve this problem.Firstly, this paper proposes a passive authentication algorithm of image region duplication forgery using merging blur and affine moment invariants. The algorithm first of all finds one merging blur and affine moment invariant which is affine invariant and highly robust for blur manipulation. Then uses this moment invariance as the feature vectors of image blocks and orders the extracted merging blur and affine moment invariant feature vector in dictionary order. The similarity degree of adjacent rows of sorted dictionary matrix by the correlation coefficient is calculated. At last, the duplication regions are located and detected by the compared threshold method. Experiments show that our algorithm is not only effective to detect and locate the region duplication forgery, but also robust to some common attacks, such as motion blur attacks and Gaussian blur attacks. Therefore, the algorithm solves the fundamental problem of image region duplication forgery with blur manipulation and has practical value.Secondly, this paper develops a passive authentication method to region duplication forgery using affine sift feature points. In this method, we first construct a complete set of affine invariance feature points, and extract the feature point descriptor of these feature points. Then match the similar feature point descriptors. At last, in order to detect and locate the region duplication forgery area, we use the threshold comparison method to find the most matching points pairs. Experiments show that, this algorithm is not only effective to detect and locate the region duplication forgery, but also robust to some common attacks, such as rotation attack, scaling attack, rotation and scaling attack, JPEG compression attack, illumination variation attack, Gaussian blur attack. In short words, this algorithm solves the fundamental problem of image region duplication forgery with rotation attack, scaling attack, and some other attacks. Thus, it can be applied to practical problems.Finally, some existing problems of our algorithms are summarized, and the future research directions are introduced.
Keywords/Search Tags:Region duplication forgery, Passive Authentication, Merging blur andaffine moment invariants, Affine Sift
PDF Full Text Request
Related items