Research On Kinds Of Detecting Reality Of JPEG Images Based On AC Coefficients Statistical Model

With the rapid development of multi-medium technology and net technology, all kinds of digital cameras and powerful function,easy operation image process software are used widely. To tamper digital images is getting easier. While it brings convenience to our life, it also makes a lot of bad impacts. For example, tampered news pictures or modificated evidential images will result in serious bad affection and disturb the justness of judiciary. It's becoming very important that how to distinguish the reality of JPEG images effectively.JPEG compress standard became the main image compress standard because of its special character. It can preserve better quality of images after a high rate of compress. So it is widely used in every fields. And it has an important signification to investigate the tampered JPEG images. This paper puts forward and improves three kinds of detection of double compressed JPEG images basis on the AC coefficients statistical model.Designing a filter algorithm for reality of JPEG image based on laplacian distribution. Obtain the eigenvectors by comparing the inosculation degree the simulative probability function of the one time compressed image's AC coefficient and the probability function of AC coefficients of the image.Depicting AC coefficients with new statistical model—balanced a steadily model. Import the definition of image complexity to cut down the calculation of the value ofα. Design a new character vector for SVM to distinguish.Aiming at double compressed JPEG images which quality factors are very close, this paper designed difference twitter mode. To detect the reality of images by minus AC coefficients which are close to and determine whether they are twittered. Experiments show that all the algorithms are effective for detecting double compressed JPEG images.