Researches On H.264/AVC Video Streaming Information Hiding Technology

With the rapid development of information technology and computer networks, the copyright protection of digital products has become an urgent issue to be tackled. It has become an urgent and realistic issue to protect the copyright and convert communication of digital video effectively. Therefore, digital video watermarking technology has been widely proposed for such an application. Up to date, most video watermarking algorithms are based on MPEG. With the development of video compression standards, since H.264/AVC standard is burgeoning video compression standard and it has outperforming compression efficiency, H.264/AVC is widely applied in video compression. Until now, the video watermarking research based on this standard is not enough.In this dissertation, compressed domain video watermarking and information hiding algorithm combined with H.264 standard's many excellent capabilities is investigated. The dissertation is organized as follows:To begin with, this dissertation briefly reviews the features and tools of H.264/AVC video coding standard, especially the application of video watermarking and its characteristics. Some typical video watermarking algorithms are reviewed.Then, a novel information hiding algorithm for H.264/AVC is proposed. This algorithm establishes a relationship between hidden messages and the carriers of middle frequency DCT coefficients by applying the classic magic matrix. Compared with the conventional one-dimensional mapping method, the proposed method based on two-dimensional mapping can obtain lower embedding cost when the utilization rate of carrier is the same. Meanwhile, the proposed method not only has minor ratio of modification, but also preserves the visual quality of the video than the existing method.Finally, a compressed-domain video watermarking algorithm robust against self-collusion attack is proposed by utilizing the new characteristics of H.264/AVC. The security of watermarking is based on the random embedding positions'selection, which is decided by the key obtained from the related change of DC coefficients. Watermark is embedded by modifying the non-zero quantization coefficients of the 4×4 blocks within a macroblock. Simulation results demonstrate that the proposed approach can enlarge the watermark capacity without sacrifice of video quality. Moreover, it is not necessary to fully decode the video stream during the watermark detection, so it can meet the requirement of blind detection with real time performance.