Abstract: For the problem of integrity authentication about compressed video streams, this paper proposed a video watermarking algorithm based on Compressive Sensing (CS). The H.264 compressed video stream was taken as the research object. By conducting random projection on the I frames of video sequences, a slight amount of measurements were obtained. Through quantization and replacement encryption, the measurements were embedded into the macro block which had the maximum amplitude in the motion vector in P frame. While authenticating, the watermark was extracted from the P frame that contains watermarked video sequences, and the similarly random CS projection on the I frames was conducted. By comparing the difference between the measurements, the integrity of video could be authenticated. The simulation results show that the intended video watermarking algorithm ensures high video quality, and little effect on the encoding rate, and it has strong robustness against attacks such as frame deletion and frame insertion.

Key words: H.264, video watermarking, Compressive Sensing (CS), random projection, video authentication

摘要： 针对压缩域视频流的完整性认证问题，提出了一种基于压缩传感（CS）的视频水印算法。以H.264压缩视频流为研究对象，通过对视频序列的I帧进行压缩传感随机投影，得到少量的测量值，经过量化和置换加密，最后以水印的方式嵌入到P帧具有最大运动矢量幅值的宏块中。认证时，从含水印视频序列的P帧提取水印，并对I帧进行相同的压缩传感随机投影，通过比较测量值的差异，实现对视频的完整性认证。仿真结果表明，该算法具有较好的视频质量，对码率的影响较小，对帧删除、帧插入、重压缩等攻击具有较强的鲁棒性。

关键词: H.264, 视频水印, 压缩传感, 随机投影, 视频认证