Blind Separation Method For Permuted Alias Image Based On JPEG Double Quantization Effect

Without any prior knowledge to verify the integrity of the digital image and automatically detect and separate the tamper area has been become an advanced research hotspot in the process of blind image forensics.The authenticity and integrity of the image played an important role in the forensic investigation,criminal investigation,information service,medical imaging and news.JPEG(Joint Photographic Experts Group)as the mainstream image compression standard in the current network environment,because it has its high compression ratio and faster transmission rate,therefore widely used in web applications and multimedia.As the image editing software functions gradually strengthened,according to their wishes to forge an image becomes more and more simple,and forge the after image just rely on the human eye cannot determine its authenticity.With the image editing software functions gradually strengthened,it's easier for people to forge an image according to their wishes,and the tampered image can't be discovered just rely on the human eyes.Therefore,the development of new technology to identify whether the image experienced a malicious tampering also becoming more and more important.At present,in the study of the JPEG tampered image,there are mainly presents the following three question.One is that most of the existing blind forensics algorithm required the compressed image at a higher compression factor,and there is no effective detection algorithm to JPEG images.The second problem is when the second quality factor is smaller than the first quality factor to the tampered JPEG image,the correct separation rate is very low.The third problem is we often need to manually set up the threshold to the binary image or a posteriori probability diagram we finally obtained.This paper made the following contributions to solve the above problems existing in the research:(1)To the gray image of JPEG format,after the Discrete Cosine Transform,the AC coefficient distribution accord with an approximation Laplace probability density distribution described.And on this basis,this paper proposed a new double compression probability model of JPEG image to describe the change of DCT coefficients' statistical properties before and after the double compression.According to Bayes criterion,using the posterior probability,the model also can show the eigenvalues of the double and single compressed block.Assign a threshold for the eigenvalues.Then the tampered region is automatically detected and extracted by using the threshold to classify the eigenvalues.The experimental results show that the method can detect and locate the tamped area effectively and it outperforms other algorithms in terms of the detection result especially when the second compression factor is smaller than the first one.(2)In real life,most of the image is the true color image of JPEG format.When an original JPEG images has undergone a localized tampering and saved again in JPEG format.The Discrete Cosine Transform coefficients of unmodified areas will undergo a double JPEG compression,and showing the double quantization effect,while the DCT coefficient of forged areas will not has this feature.We get the posterior probability density of the tampering image.Assign an adaptive threshold for the eigenvalues with the Particle Swarm Optimization.Then the tampered region is automatically detected and separated by using the threshold to classify the eigenvalues.Compared with other traditional methods,the proposed approach could effectively separate the tampered regions from the tampered image without respect to the location,size and number of tampered image.The research work for blind image forensics provides a new thinking method and theory.The experimental results show that it outperforms other traditional algorithm algorithms.Study on the algorithm of this subject has a certain practical value.