Research Of Image Blind Tamper Detection Algorithms

In recent years,with the popularity of digital images in the media.Image becomes the channel where people are getting more and more important.But such as the South China tiger event similar to the endless image of forging events,triggering people on how to identify the authenticity of the image of thinking.Digital image tamper detection technology is in this context came into being.Mainly through digital image processing technology to study how to identify the image is really false technology.This dissertation first elaborates the background and significance of the field of tamper detection.And then a more comprehensive introduction and comparison of the current two major detection algorithm category,that is,active detection and passive detection of various types of algorithms and the advantages and disadvantages.This dissertation introduces the related technology of passive tamper detection.And for the passive tamper detection technology,the two most important means of tampering,with the picture of the copy-paste tampering and different pictures of the tampering proposed two improved detection program:(1)copy-move tamper detection based on HSV and SIFT.Mainly for the same picture of the copy-paste tamper detection.Based on the study of various copy and paste tamper detection algorithms,a new algorithm is proposed to extract the HSV feature of the image as a SIFT feature for the neglect of the color feature of the traditional SIFT algorithm.Firstly,the SIFT algorithm is used to find the feature points and SIFT eigenvectors,and then HSV eigenvectors are extracted by HSV mean of feature points.Finally,the 128-dimensional SIFT feature vector and the three-dimensional HSV feature vector are matched to determine the tampering area.(2)based on multi-domain feature fusion splicing tamper detection.Mainly for different pictures of tampering tamper detection.A new algorithm based on three domain feature fusion is proposed for the natural distribution of natural images in spatial domain,DCT domain and DWT domain.In the airspace,the Gaussian model of the natural image is used in the DCT domain,and the subband feature is used in the DWT domain.A new feature extraction scheme is designed and the splicing image is detected by SVM classification model.Experiments are proved to have better effect than the detection algorithm based on single domain.Experiments show that the two detection algorithms proposed in this dissertation are more optimized than the previous general detection algorithms.Based on HSV and SIFT copy-move detection,the detection of images with similar backgrounds is higher.Patch detection based on multi-domain feature fusion is also better than the detection algorithm based on a single domain feature.The two detection algorithms proposed in this dissertation cover the two most important image tampering methods for copy-move tamper detection and tamper tamper detection.Can be compared with the traditional detection algorithms have different degrees of optimization.