Composite Image Forensics Research Based On Edge Detection

With the rapid development and further improvement of computer and multimedia technology, our society has entered the era of image-centered digital culture, however, due to the extensive use of Photoshop, ACDSee and other professional image processing software, the tampering of digital images also become more and more easier. A variety of intentional or unintentional image tampering may have serious impact on society, therefore, a new technology is needed to protect the authenticity of digital images.Digital image forensics is divided into two categories:active forensics and passive forensics. Active forensics technology of digital image requires to embed additional information such as watermark into images in advance, therefore, this approach has significant limitations.Passive forensics of digital image breaks these limitations, it is also known as blind forensics. Based on the inherent properties of digital images, passive forensics can identify the aboriginality, the integrity and the authenticity of image content, without embedding additional information such as watermark in advance.This thesis systematically summarizes the research base and the development direction of blind forensics, we primarily study on the commonly used technology of digital image tampering. According to the principle of tampering approach, in this thesis we propose a corresponding approach based on the traces left in the tampered image, this approach can identify if the image is tampered, and can locate the tampered regions for the tampered image. The technological innovation and research results are as follows:Firstly, based on the study of composite images that have been blurred, we research the impact of blurring on the digital image grayscale standard.Secondly, as the linear scale-space theory can well reflect the idea of multiscale analysis, we analysis the impact of the scale factor in linear scale-space on image edge blurring, which can provide theoretical basis for identifying images and locating tampered regions in tampered images.Thirdly, we apply the edge detection operator to digital image forensics. As the blurring reduces the contrast of data, the difficulty of detecting image target is increased, and the precision of target recognition is reduced. Therefore, we need to use edge detection operators to lock the edge of tampered target. Experimental results demonstrate that the proposed approach can accurately detect the blurring traces of composite tampered images, and can precisely locate the tampered region.Finally, we summarize our research work for the thesis, and discuss some research topics and directions relative to this work in the future.