Research On Key Technology Of Copy-move Forgery Blind Forensics For Digital Images

With the increasing popularity of digital image acquisition equipment,image becomes an important way for people to get news and information in daily life.At the same time,with the development of photo editing tools,such as Photo Shop,more and more people could easily edit the image,such as splicing synthesis,background rendering,region duplication,etc.,which brings a variety of fun for people's life.But in recent years,in the news reports,academic research,court evidence and other places that require completely true images,some people abuse editing tampered images to achieve selfish purpose,where serious cases have violated the law.Based on this,image forgery forensics technology has become a hot topic in recent years.In this paper,copy-move forgery,which is the most common means of image forgery,is studied.The specific content is as follows:1.Copy-move forgery detection algorithm based on color LBPThe existing pre-processsing of copy-move forgey detection algorithms for color image include two ways: ?1 color image is converted to grayscale image,which completely discard the image of the color information and reduce the forgery detection accuracy;?2 three channels of color images are calculated respectively,where the calculation is three times of single channel and greatly increase the algorithm time-consuming.Based on this,this paper presents a novel copy-move forgery detection algorithm based on color LBP(Color Local Binary Patterns)image and improved kd tree hyperplane partitioning split search.Firstly,the color image is preprocessed,that is,the color LBP texture image is established to fuse color information and LBP texture feature.Secondly,the Gray Level Co-occurrence Matrix(GLCM)features are extracted.Lastly,the improved kd Tree and hyperplane partitioning split search method would quickly match image blocks,and applying morphological operations to remove mismatches.The experimental results show that the proposed algorithm is effectivefor covert tampering,and has high robustness to blur,noise and JPEG recompression.2.Copy-move forgery detection algorithm based on scaled ORBBinary features have the characteristics of fast extraction and high matching accuracy,which is a popular feature descriptor in recent years.Among them,ORB(Oriented FAST and Rotated BRIEF)has the functions of translation,rotation,distortion invariance,but no scale invariance,which causes the robustness of the copy-move forgery with scaling operation is poor.Based on the problem of false matching and poor robustness,this paper proposes a copy-move forgery detection algorithm based on scaled ORB.Firstly,the Gaussian scale space is established.Secondly,the o FAST(oriented FAST)points and ORB features are extracted from the scale space respectively.Thirdly,the o FAST points are mapped to the initial image and scaled ORB features are matched according to the Hamming distance.Finally,RANdom SAmple Consensus(RANSAC)algorithm removes the false matching pairs and locates tampering areas.The experimental results show that the proposed algorithm could not only resist geometric transformations,such as scale transformation and rotation,but also has high robustness to postprocessing,such as blur,noise and JPEG recompression.3.The blind detection algorithm distinguishing the SGO real image and copy-move forged imageNatural scenes often contain two or more similar but genuine objects(SGO),such as trademarks,buildings,etc.,which brings a huge challenge to traditional copymove forgery detection.The existing copy-move image database only contains geometric transformations,such as rotation,scale,and post-processing operations,such as noise,blur,and JPEG recompression.There is no discussion about the existence of SGO.Based on this,this paper first presents a new Copy-move forgery database(COVERAGE),which contains 100 SGO real images and their corresponding copy-move forged images.What's more,it uses 6 kinds of tampering operations,such as translation,scaling,rotation,free transform,illumination transformation(linear and nonlinear illumination transform),and combined transformation.Based on the SGO real image and copy-move forged image,we found that the similar regions of the SGO real image do not have the affine transformation matrix.Based on this,this chapter proposes to distinguish SGO real image and copy-move tampering image Blind identification algorithm.4.Copy-move forgery detection algorithm for illumination transformationTo solve the problem of poor robustness on the illumination transformation,a new method based on Maximally Stable Extremal Regions(MSERs)and Localized Intensity Order Pattern(LIOP)is proposed.LIOP feature exhibits a good invariance to linear illumination transformation,but has poor robustness to complex illumination transformations.Therefore,for the non-linear illumination transformation operation,this paper proposes a copy-move forgery detection algorithm based on the Difference of Gaussian(DOG)and Mixed Intensity Order Pattern(MIOP).The MIOP is the fusion of Overall Intensity Order Pattern(OIOP)and LIOP,which use independent oriented sampling strategy in the local feature calculation.It has geometric rotation invariance and good robustness to the complex illumination transformation.