Research On Passive Forensics Method Based On Digital Image Tampering

Digital image forensics is a technique which is used to detect the changes of digital images. It is widely used in the authenticity identification and the court evidence. Digital image forensics technology is to detect changes, find out the place where the artificial modification, and then mark the place of modification. Compared to the digital image active forensics need pre embedding information, the digital image passive forensics is in not pre add password authentication, content of digital images can be artificially modified differential method for image verification. Its application is more extensive and have more high research value. In recent years, with the rapid development of software, the function is much stronger. However, the digital image passive forensics does not match with the development speed, so that the lag is also taken for granted. Because, only the emergence of new means of tampering, will make people to study how to detect tampering with the place. Such development is also in line with the basic process of research, found the problem and then to solve the problem. Such a situation, exposed the forensic technology have strong pertinence, one to one of the detection efficiency is low. At the same time, the accuracy of detection has been now the soft underbelly of detection technology, in a variety of tampering with the means of comprehensive application in digital images, the accuracy of a single detection algorithm to detect really is not satisfactory.According to digital image change detection research, relates to copy-paste tampering of digital image passive forensics and JPEG image tampering forensics methods, the existing algorithm experiments were conducted analysis and summary. According to the deficiencies of the existing methods proposed two kinds of improved digital image passive forensics algorithm, research on copy-paste tampering using LPP and based on the random sampling of the maximum likelihood estimation of JPEG image passive forensics method. The work of this paper is mainly based on the following two research:The blocks of image feature extraction and dimensionality reduction processing have been investigated for the proposed locality preserving projection algorithm of feature vector of dimensionality reduction, the goal is to find a transformation matrix P. will be in the high dimensional space data set is mapped to spatial position. Finally, this paper put forward a improved enhanced cross difference filter, to strengthen the block effect and reduce the interference from the edge of the image and texture and integration of random sampling, voting two methods to improve the maximum likelihood method, get a new estimation method. which is based on random sampling maximum likelihood estimation to improve the accuracy of the block size estimation (RSMLE).