Feature Fusion Based Recompression And Resampling Forensics Of Digital Image

The digital image processing software,which is highly intelligent and easy to operate,brings great convenience to people's daily life,and also causes a series of latent information security problems,and a reliable digital image forensics method is urgently needed to maintain social order and fairness.On the principle level of image processing,recompression operation must be done during image processing;in the content modification level,resampling operation is the image processing process common operation,including the magnification,the reduction,the rotation and so on,these two kinds of operation above can bring the comprehensive auxiliary test result for the image forensics.This paper systematically sums up the research foundation and development direction of digital image double compress and sampling,respectively from evident framework of the feature fusion recompression and resampling based on feature fusion,JPGE recompression detection based on feature fusion,and JPEG resampling detection of based on feature fusion,and the development of tampering-forensics system based on feature fusion of image recompression and sampling were studied form these four aspects.The main work of this paper is as follows:(1)The paper studies the forensic framework of recompression and resampling with feature fusion.In this paper,the existing detection algorithms at home and abroad are theoretically analyzed and summarized,for the forensics areas of recompression and resampling of digital image,there are three deficiencies in all exist methods,including no standard evidence frame,a single forensic feature,the large dimension of the eigenvector.On this basis,this paper is from the point of view of feature fusion,chooses the advantages of each complementary feature,seeks the optimal feature fusion algorithm,assists the appropriate machine learning methods,and designs a recompression forensics framework based on feature fusion,a resampling forensics framework based on feature fusion and a unified forensics framework.The principle and specific algorithm of following in the framework are teased,including CCA and DCA which based on the feature fusion method mentioned,PCA,Prob-PCA,SPE,and Sammon mapping method based on feature parallel fusion and dimensionality reduction,SVM,gcForest,CNN of machine learning.(2)Feature fusion based JPEG image recompression tampering forensics.According to the design of recompression tampering-forensics framework based on feature fusion,in this paper,the first effective digital feature of the DCT coefficients that describes the influence of the Y-channel compression and the difference of adjacent coefficients between the DCT coefficients affected by the recompression operation and its surrounding coefficients are presented to detect the digital image recompression.The validity of the algorithm is proved by the comparison experiment method.The method solves the problem that the detection effect is poor with the diagonal and the second compression factor of 95,the redundant data with the feature vector is too large,and the data quantity is too big to affect the detection efficiency.(3)Feature fusion based JPEG image resampling tampering forensics.According to the design of the resampling forensics frame,the paper presents the texture characteristics of the local periodic correlation which is affected by the resampling operation,describes the Benford characteristics of differences in the effects of resampling operations on R,G,B three channels,and describes the adjacent coefficient difference features of the relationship between the DCT coefficients affected by the resampling operation and their surrounding coefficients,and describes the block effect features of the double JPEG sampling,and these four kinds of feature are proceed feature fusion operation of digital image resampling detection,which is proved by contrast experiment.This method solves the problem of detecting the insensitive and irrelevant characteristics affecting the detection accuracy and the large amount of data affecting the detection efficiency when the sampling factor is near 1.0 in the recompression forensics.(4)The development of image recompression and resampling forensics system based on feature fusion.Based on the unified forensics framework,this paper designs and develops an automatic forensics system based on feature fusion for digital image recompression and resampling,and proves that the system is simple,interactive,feasible and portable by scientific test method.(5)Finally,the paper is summarized and the prospect of the future research and developing direction of the digital image recompression and resampling forensics technology is also put forward.