Research On Detecting Double JPEG Compressed Image With The Same Quantization Matrix

With the development of information technology and the widespread spread of image acquisition equipment,digital images have become important media forms.At the same time,the functions of various digital image processing software are becoming more and more powerful,and the use-pattern is becoming simpler and simpler.In recent years,more and more literatures have been researched on digital image forensics related technologies.Compressed history analysis can be used as one of the bases of digital image forensics and is an important part of digital image forensics.Since JPEG images are generally saved as JPEG format again after tampering,tampered JPEG images tend to undergo compression twice.Therefore,the detection of JPEG recompression can be used as an important basis to judge whether the image has been tampered.The research work of this paper is as follows:(1)In order to improve the performance at low quality compression,an algorithm based on spherical coordinates to detect double compressed JPEG images is proposed.Conversion error is proposed by analyzing the effect of JPEG compression on color space conversion in the proposed method.The truncation error and rounding error are classified at the pixel level and mapped into spherical coordinates which can avoid the occurring of mixing error to obtain more obvious features.Comparative experiments show that compared with the previous algorithm,the accuracy rate of this paper is improved by 5% on average in the case of low quality factor,and the accuracy rate reaches 99% if the quality factor is 80 and 85.(2)In order to improve the performance at cross detection experiment,an algorithm based on image to detect double compressed JPEG images is proposed.The proposed method analyzes the characteristics of JPEG images.If the image is compressed with the same quantization matrix many times,the image tends to be stable,which can obtain convergence error.Transposition attacks can interfere with the stabilization process to obtain transposition errors.Hyper-parameter mapping is used to reduce the redundant information in the convergence error and expand the impact of transposition attacks.Comparative experiments show that the scheme can not only achieve better performance,but also approach the theoretical upper limit.In addition,compared with the previous algorithm,the performance of the cross detection experiment has been greatly improved.