Research On Image Copy-paste Detection And Sharpening Forensics

The growth of digital images,digital images play an important role in many fields with their rich information and intuitive performance.Such as medical treatment,news media,court trials,etc.At the same time,the development of image editing software allows ordinary users to easily modify the image content to achieve the purpose of use.This brings great challenges to the authenticity and originality of images,and digital image forensics has become an important and urgent research topic.This article mainly focuses on the detection methods of copy and paste tampering and sharpening tampering in the field of passive forensics of digital images.It realizes the detection of conventional texture image tampering and the comparative analysis between algorithms,and further discovers and solves the main problems of texture smooth tampering area detection.Simultaneously,an effective USM(Unsharp Mask)sharpening detection method for small-size images is realized,and experiments have verified that this method is superior to existing methods in the sharpening detection of small-size images.The following are the main research tasks:First,by analyzing the performance of existing copy-and-paste tampering detection methods,select a tampering detection method based on key points.First,the detection method based on Harris corner point extraction combined with LCP(Local Configuration Pattern)feature description is adopted,and the copy and paste detection of the conventional texture region tampered map is realized by using G2 NN feature matching and RANSAC false matching elimination.Then,so as to settle a matter of low detection rate after scaling,rotating,and compressing the copied area in the previous method,the SIFT(Scale Invariant Feature Transform)key point detection method is adopted,and the texture feature in the neighborhood of the key point is added to describe the change of the LBP(Local Binary Pattern)processing small pixel.The comparative experiment proves that the algorithm has better robustness.Then,in order to solve the problem that the existing copy and paste detection algorithm based on key points cannot detect the tampering of the smooth region of the texture,the UR-SIFT(Uniform Robust Scale Invariant Feature Transform)key point extraction method is adopted to realize the uniform distribution of key points in the image space and scale space.The standard SIFT algorithm relies on the global threshold to filter the defects of the feature,and the Opponent SIFT is used to enhance the recognition ability of the feature descriptor.Then,a two-stage feature matching algorithm is used to obtain a more accurate matching result.Finally,by estimating the affine transformation experienced by the tampering area,compare the similarity before and after transformation,the tampering area is located.Finally,aiming at the problem that a sharpening operation is usually added to hide modified imprint in the post-processing of image tampering,an effective small-size image unsharp mask sharpening detection algorithm is realized through the research and analysis of existing sharpening forensics methods.Take the Y component of the color image YCr Cb color space as the initial feature extraction image,use the differential statistical histogram of the Y channel block DCT(Discrete Cosine Transform)transformation image to achieve sharp feature extraction,and finally use SVM to train and classify the extracted feature vectors to achieve sharpness decision making.