| With the development and growth of the computer and the Internet,various streaming media aggregation platforms,such as live video,video on demand and Internet TV have entered people's lives.Video resources have already been the lifeblood of all major platforms,but accompanied by information security issues.Various infringements,tampering,and copyright issues have had a significant impact on the healthy dissemination of video resources,causing great losses to various video aggregation platforms.The information security issue needs to be resolved immediately.The emergence of digital video watermark provides a more effective solution.However,due to the complexity and particularity of video signals and digital video watermark not only should have the basic characteristics of real-time processing and imperceptibility,but also should adapt to the actual application scenarios,In the live video application,there is high requirements for the real-time performance and video quality of the video transmission.On the basis of ensuring invisibility after embedding the watermark,the video quality and the real-time performance of the video transmission should be less affected.The video should not be attacked by video coding compression,video frame scrambling,frame recombination and frame loss etc.It should be qualified for blind extraction and random detection and embedding video in a real-time manner.Moreover,it can be correctly extracted after the platform processing and improve the anti-attack ability of the watermark.In this paper,the relevant theoretical technology of digital video watermark is combined with the compression technology of H.264/AVC video coding standard to carry out the following work:This paper studies the related technology of video watermark and analyzes the classic video watermark algorithm.The paper adopts the video watermark algorithm of DCT(Discrete Cosine Transform)coefficient of video region.The core idea is to divide the video frame into regions and distinguish the energy value of different blocks by DCT coefficient.The watermark embedding region is a region where the DCT coefficient is not zero.And at the same time,in combination with the motion vector of the block,a block with a larger motion vector value is selected for watermark embedding.The algorithm considers both intra-frame and inter-frame information,and classifies video sequences according to motion vectors and block complexity.Within the frame,4×4 image blocks are classified according to whether they contain detailed information and according to the different visual characteristics of the human eye on the stationary object and the moving object.The motion detection is performed between adjacent frames,and the image is divided into fast motion block and slow motion block.Through the two-layer detection mechanism,the intersection of the block containing the detailed information and moving faster is selected to embed the watermark,so that the position of watermark can be adapted to the characteristics of the human visual model and video signal.And in the meanwhile,the embedding strength can be adapted to different detailed information of the corresponding block.Moreover,the algorithm also applies the spread spectrum technology and random scrambling technology to pre-process watermark before embedding,so that the embedded watermark has better robustness and security.The experiment uses open source webRtc to verify algorithm and adopts video watermark evaluation method to evaluate.The results show that the algorithm can resist common attacks such as re-coding,frame recombination,frame deletion etc,and can quickly embed and extract watermark information,which improves the efficiency of watermark embedding and reduces the errors caused by quantization and watermarking.It is applicable for application scenarios of live video and video on demand copyright protection to ensure information security. |
PDF Full Text Request