Research On Passive-blind Detection Algorithm Of Digital Image

The detection and judgment of authenticity of digital image has become one of theimportant researches in the area of digital image automated processing in recent years.It has been used widely in many aspects in social life, such as judicial evidence,anti-counterfeiting of trademark, medical image and so on. Active image detection andpassive image detection are two ways of the algorithm of digital image detection. Activedetection needs to be inserted identifying information before detection and identifiesauthenticity or distorting of image by detecting the information. Passive detection doesnot need to be imbedded information before detection and only detects digital imageaccording to all kinds of feature information in the image itself. Passive detectionalgorithms have been becoming more popular because information does not need to beembedded before processing. So, more and more researchers are concentrating on thisdirection.In the thesis, it focuses on the passive-blind detection algorithm of digital image. Weintroduce the background, practical significance, research status at home and abroad anddifficult points of the algorithm in detail. The main research contents and the results areas follows:1) Some usual detection algorithms are introduced. Those algorithms are based onthe characteristics of the image itself, especially the most classical algorithm whichidentifies the authenticity of image according to the block matching. Many detectionmodes appear based on the algorithm above, such as exhaust law of searching, matchingbased on blur moment invariant, matching based on singular value decomposition(SVD)and matching based on principal component analysis(PCA). On this basis,algorithms add the module of transfer vector statistics in order to locate the tamperingarea in digital image more precisely.2) We introduce the passive detection algorithms of JPEG format which is the mostcommon in daily life. In the thesis, we talk about the generation process of JPEGformat in detail and analyze the specificity of passive detection on this kind of format.In this thesis, we introduce three detection algorithms specially aimed at JPEG format,including algorithm based on block artificial grid(BAG) mismatching, algorithm ofquantization table estimation in JPEG format based on the noise inconsistency in imagebackground and algorithm based on periodicity of double compression. 3) After comparison of three algorithms and analyzing the advantages and defectswe draw a conclusion that the algorithm based on BAG mismatching is better because ofthe excellent performance. But there is a major defect which usually mistakes originalarea in image for distorted area and produces many errors. If the value of quality factoris higher or drastic change of gray value gradient exists in the digital image, the falsealarm rate usually becomes higher and seriously impacts on detection results. But froma large number of trials we can find that there is an indiscoverable difference betweenfalse alarm area and tampering area in test image. There are two adjacent boundariesbetween mismatching BAG in both vertical and horizontal directions in tampering area,however, the type of phenomenon does not exist in false alarm area. Finally, we analyzethe causes of the difference in two areas from the original algorithm.4) In the thesis, we propose a kind of improved way according to the differencebetween original area and distortion area. After testing the pictures in digital imagelibrary, the statistical data demonstrates that the modified algorithm exhibits betterperformance than that of the original algorithm in several tests.