Research On Content-based Image Tampering Detection Method

With the rapid development of the digital image processing technology and the widespread use of various intelligent devices,the application of large-scale image/video database has become ubiquitous.At the same time,the emergence of many powerful and easy-to-operate image processing software,even the non-professional users can easily manipulate and modify image content without leaving traces of operation.The manipulated images are widely spread on the Internet,enriching people's visual experience,but they have also caused extremely harsh negative effects in various fields of society.Therefore,in recent years,digital image tampering detection technology has become new research focus in the field of information security,which has important theoretical significance and practical application value.This paper mainly studies image hash technology based on active forensics and image splicing detection technology based on passive forensics,the main work is as follows:As an active forensics technology,image hash has important application in image content authenticity detection and integrity authentication.We studied the image hash generation method,propose a perceptual image hash generation method based on hybrid features and a hash-based image content tampering forensics method.In the proposed method,we use the color features of the image as the global features,use point-based features and block-based features as local features,and combine with the structural features to generate intermediate hash code.Then we encrypt and randomize to generate the final hash code.Based on the generated perceptual image hash,we construct a coarse-to-fine grained method for image content authenticity and integrity forensics,which can realize image content tampering detection and tampering location The proposed method can realize object-level tampering detection and tampering localization.Abundant experimental results show that the proposed method is sensitive to content changes caused by malicious attacks,and the tampering localization accuracy achieves pixel level,and is robust to geometric distortion and content-preserving manipulations.Compared with state-of-the-art schemes,the proposed scheme with more superior performance.As another image content authenticity forensics technology,passive forensics is a technology for forensic analysis of image contents by using the internal features of the image itself without relying on any prior information.We propose a coarse-to-fine grained image splicing localization method.Considering that photos from different cameras generally carry the camera's own pattern noise and the noise introduced in the imaging procedure,generally speaking,the noise level of the splicing image region is inconsistent with that of the original image region.Based on this fact,we carry out the splicing forensics by looking for regions with inconsistent noise levels in the image.In the proposed method,we used Laplace operator to extract the local noise of the image,and estimated the local correlation feature by using the prediction residual.Then,we use the clustering algorithm to cluster the forensic features and get suspicious spliced regions.Finally,we refine the suspicious spliced regions to get the precise splicing regions.Compared with the existing noise-based image splicing localization methods,the proposed method has high detection accuracy and stronger robustness for a variety of content-preserving manipulations.