Blind Detection Of Digital Images Tampering

The popularity of Digital imaging equipment and image editing tools bring the convenience in our life, but also exploited by malicious tampering, brought many negative issues. These malicious tampering threat to the authenticity of the press, judicial impartiality, and the reliability of scientific research, while the right of the individual portrait and other rights are also seriously violated. Frequent occurrence of malicious tampering incidents confirmed that digital image forensic is very important. Image forensics has become a focus in the field of information security and forensic research. Traditional active image forensics'technologies are all add prior information into the image such as digital watermarking, digital signatures. However, this method is now lack of generality when numerous generate and rapid dissemination of digital images, the scope of useful are also become small. Therefore, the study of natural image forensics technology is the key point to solve the current a large number of image forensics. The main contribution of this dissertation is summarized as follows:At first, this paper analyzes the existing image forensics technologies'principle, characteristics of the existing theoretical and study the background of forensic technology.Secondly, the paper proposes the method of detecting image splicing. In this section, the tampering was divided into copy-move (in an image) and copy-paste (in two images). The first one is copy a part of the image and paste on another part of this image in order to hide the important goals. The other is achieved by two or multiple digital images by copying a portion in one image and paste in another image in order to create a false impression. ThenThirdly, propose a detection algorithm to detect the image copy-move forgery based on extracting feature from wavelet sub-band. In order to further reduce the computation, we extracted the gray distribution feature from wavelet sub-band, iterative division method combined with the use of similarity are used to searches similar image block.Fourthly, an image splicing detection scheme is proposed. The scheme is based on image quality and analysis of variance. Four kinds of noise used to simulated the tampering of image, and extracted the changes of image quality changes which caused by image splicing, and analysis of variance is used to selected the image quality measures which are more sensitive to image blind splicing detection. Combined with the characteristic function moments of three-level wavelet sub-bands and the further decomposition coefficients of the first scale diagonal sub-band, we extracted all features from given image and its predicted error image. SVM is adopted as the classifier to train and test the given image.At last, the higher order statistical moments of wavelet package was used as characteristics to detect for splicing image. We use principal component analysis (PCA) to reduce the dimension of detection feature. This method also can get better performance.