Copy And Move Of Color Digital Image Based On Feature Research On Tampering Blind Forensics Technology

With the rapid development of technology in the fields of semiconductors,sensors,and computers,digital image technology has also made great progress.Digital images and videos enter every aspect of people’s lives,work,security and forensics.Images and videos have become an important basis for seeking the truth on many occasions.Accompanied by this,the tampering technologies for images and videos are constantly updated and varied,and the traditional thinking of "there are pictures and the truth" is constantly challenged.Therefore,the identification technology of tampered images becomes more and more important,and it has engineering application value to scientifically study how to tamper with digital image content and verify the authenticity of image content.The research of this paper mainly focuses on the blind forensic technology of color digital image copy and paste.The main work of this paper includes the following points:(1)In-depth analysis of digital image copy-paste tampering blind forensics technology,the image features used in this algorithm are introduced in detail,and the problems existing in digital image copy-and-paste tampering blind forensics algorithm are analyzed:the tampered area of self-similar images rotates at a specific angle The problem of low accuracy of GLCM feature detection and the problem of low accuracy of HOG feature detection when the tampered area of the non-self-similar image is rotated at any angle.(2)In view of the low accuracy of traditional algorithms in detecting the rotation of self similar images at specific angles,the traditional LBP feature extraction is improved:the LBP feature of color digital images is used to process the norm of three primary color values,extract clearer texture details(groove depth,density,thickness,etc.)and generate LBP images,So that it can tamper with and obtain evidence on self similar images(images with slow gray value transformation of pixels and their surrounding spatial neighborhood).The statistical method of GLCM features is improved:the eight directions and two types of pixel step information of the image block are counted.Because the GLCM values are equal in the eight directions,the rotation tampering area with a specific angle can be detected.Based on this,a copy and paste blind forensics algorithm of tampered image with GLCM feature is designed and implemented.(3)In order to solve the problem that the detection effect of non self similar images for false images rotating at any angle is not good,the hog feature extraction method is improved:this method sets a moving rotation region in the image block,calculates the gradient value and gradient direction value for the moving region,and the gradient value and gradient direction value of each moving region remain unchanged.This improvement makes the hog feature rotation invariant at a certain angle.In order to increase the correct detection rate of tampered areas rotated at any angle,the fusion feature vectors of hog and Zernike moments are extracted from the dense orb key point image block,and the tampered images are matched for forensics.The fusion method can detect the tampered areas rotated at any angle.Based on this,a tamper blind forensics algorithm based on the double feature fusion of hog and Zernike moment is designed and implemented.The algorithm is tested and compared by using test sample database and evaluation index,and the performance of the algorithm is evaluated.The experimental results show that the image tampering blind forensics algorithm based on GLCM feature and the image tampering blind forensics algorithm based on the dual feature fusion of hog and Zernike moment have certain application value in the field of image falsification.