Research Of Detecting Algorithm For Double JPEG Compression With The Same Quality Factor

Digital image forensics is an important research direction in the field of multimedia information security,for the research of digital image forensics that based on image format,mainly aims at the JPEG image compression.Double JPEG compression is an image that experienced twice JPEG compression process successively,image manipulation often lead to double JPEG compression,and some image steganographic schemes might generate double JPEG compressed images.Therefore,the double JPEG compressed image detection research is particularly important in the digital image forensics.According to whether the quality factor of the primary and secondary JPEG compression are identical or not,double JPEG compression can be categorized into two classes: double JPEG compression with different quality factors and double JPEG compression with the same quality factor.At present,detecting research for double JPEG compression with different quality factors has achieved some successes,many successful detection algorithms have been proposed,however,for the double JPEG compression with the same quality factor,is still in the primary research phase,detection accuracy of detection algorithm is low to same extent.In this dissertation,we analysis the error that exist in the process of JPEG compression,proposed two improved detection algorithms which based on error image,detecting double JPEG compression with the same quality factor,in order to improve the detection accuracy.The main research contents include:(1)Analysis the error that exist in the process of JPEG compression and decompression,and then extract the error image from JPEG image to be detected,using the extracted error image instead of JPEG image,thus proposed the detection algorithm of double JPEG compression.(2)After successive JPEG compression,analysis the statistical difference of data of the error image,at the same time,analysis the statistical difference of data of the error image that under different quality factor.Different from other detection algorithms extracted features directly from the JPEG image,in this dissertation,the two improved detection algorithms proposed in the dissertation extract a set of feature vector from the error image of the JPEG image,used for distinguish the single JPEG compression image and the double JPEG compression image with the same quality factor.(3)Using the Support Vector Machine,test the feature vector that extracted from the two improved detection algorithms,detecting the single JPEG compression,and the double JPEG compression with the same quality factor,detecting the triple JPEG compression with the same quality factor as well as,at the same time,performance comparison has been carried out with two existing detection algorithm.The experimental results show that under different compression quality,for the two improved detection algorithms proposed in this dissertation,the detection accuracy generally higher than the accuracy of existing detection algorithms.At the same time,A double JPEG tamper detection framework is designed,the tamper detection of double JPEG compression image with tamper is completed,realize tamper detection and tamper location of double JPEG compression image with tamper with the same quality factor.