Research On Video Watermarking Technology In H.264Compressed Domain

In recent years, due to the emergence of more and more digital products, the issue of copyright protection is paid more attention. The digital watermarking technology as an effective means becomes one of research focus. The latest H.264standard becomes the most common one for video compression code as its excellent compression ration and better network affinity. For a low bit rate video, such as H.264, its better compression performance brings the large challenge for the video watermarking technology. Hence, in this paper, we deeply studied the video watermarking technology based on H.264compression domain.To overcome the shortcoming of the small key space of the tradition Arnold scrambling technology, some improvements for scrambling are presented in this thesis. Firstly, the watermarking image is divided into four parts. Then, each part is scrambled separately. At last, the overall synthesized image was scrambled. Additionally, in order to improve the key space further, more modified is proposed, the chaos theory is adopted to combine with the improved scrambling theory mentioned above. This double encryption method for watermarking information improves the key space further. Therefore, the security of watermark information was insured. It is ready for a new video watermarking algorithm.A new video watermarking algorithm based on H.264compressed domain is proposed. To increase the robustness, the4×4blocks which have more nonzero quantized coefficients are chosen as embedding locations by analyzing the texture complexity of macroblock. On the basis of the distribution characteristics of DCT coefficients, the probability matrix is determined by utilizing the variance of DCT coefficients in different positions under the different intrapredicted models. Quantized coefficients of the luma components in I frames is modified based on the probability matrix. So the watermark is embedded by changing the odevity of the number of the nonzero quantized coefficients. At the same time, the embedded watermark can be extracted by utilizing the odevity of the number of nonzero quantized coefficients of the block.Some experiments of digital watermark embedding are simulated by utilizing the proposed method for different video sequences. The performance of the watermarking is evaluated by calculating several parameters such as the peak signal-to-noise ratio, extraction rate of watermarking, bit-rate increase and the encoding time and so on. Simulation results and analyzing show that the proposed algorithm has good invisibility and real-timing, and can resist some common attacks. In addition, there is no increase on the bit-rate of video after embedding the watermarking. Compared to other algorithms, the proposed method can prevent bit-rate increase and improve embedding capacity without sacrificing perceptual quality. The algorithm can also meet the requirements of blind detection in extracting process.