Research On Digital Video Forensics Algorithm Based On Optical Flow And Watermark

With the continuous progress of the science and technology and the rapid development of digital media, digital video has become an indispensable record tool in people's daily life. In this day and age, there are a large number of digital videos on the Internet, which contain a variety of content. However, with the appearance of video editing software, it has become easily for people to tamper a video without leaving visual tracks. These digital videos which has been tampered will influence the fairness of the law and the order of the society. As a result, ensuring the video's primitiveness and authenticity has always been a popular topic which has academic significance and practical value.Based on the difficulties mentioned above, in this paper, we pay our attention to the research on digital video forensics. Two kinds of digital video forensic algorithms are proposed which belong to active authentication and passive authentication respectively. The one is digital fragile watermarking algorithm based on inverse number, and the other one is video inter-frame forgery detection algorithm based on the optical flow consistency. The main efforts in our paper is as follows:As for digital videos, we proposed a method that hiding the secret information without any awareness in video frames based on the inverse number which is a basic concept in linear algebra, and make the video frames distortion as small as possible. Thus identifying the authenticity and integrity of digital video content.Moreover, another novel detection method based on Horn-Schunck(HS) optical flow algorithm is proposed, which firstly convert all frames to gray images and assumes the global smoothness constraint and smooth every frame. Secondly, it estimate the three partial derivatives of every frame brightness. Finally, it relies on iterations to figure out the optical flow variation sequences and the magnitude of optical flow sequences. This method relies on the fact that forgery operation will introduce discontinuity points to the optical flow variation sequence and the magnitude of optical flow sequences. By contrast, the proposed method can improved the detection efficiency immensely while ensuring the algorithm reliability.