Study In Video Watermarking Based On MPEG-4 And H323

With the development of computer networks and multimedia techniques, the copyright protection and information security becomes a hot topic. A common method is imbedding information to the original information, so that we can judge the ownership properly. This is so call watermarking technique. Digital watermarking has first been extensively studied for still images. Today however, many new watermarking schemes are proposed for other types of digital multimedia data, the so-called new objects: audio, video, text and 3D meshes. As one part of a key program supported by National Natural Science Foundation of China "Research on theories and techniques of security of image and graphics under network circumstances " (60133020), some issues encountered in digital video watermarking based on MPEG-4 and OpenH323 are presented and discussed in this paper.First of all, we cover some basic topics about digital video watermarking, including the difference between video watermarking and image watermarking, the research status in quo, the applications, characteristics, challenges and trends in video watermarking.Secondly, after introducing the verification model of MPEG-4 simply, we discuss the MEPG-4 video watermarking technology. Then we devote ourselves to the following two aspects:(1) Geometric transformation is one of main obstacles to video object watermarking in that it results in the loss of synchronization during watermark detection. Image normalization is used to achieve synchronization by comparing the normalization info before watermark embedding/detection. Then the fluctuating points caused by interpolation when normalizing is adjusted. Also in order to make the watermark be more robust, we embed the watermark in the coefficients of SA-DCT.(2) The DCT coefficients in the I frame of video object may be few in that the video object may be very small, so we study the watermarking of motion vectors in P frame and B frame. We adopt the scramble arithmetic based on life game to get the embedding positions, so as to have good resistance against Interpretation Attack. Then we embed watermark through modifying the motion vector type flag.Lastly, some preliminary researches are also taken toward video communication. Because that video communication calls for real-time, blind-detectable, and robust watermarking, the following strategies are put forward: firstly encoding message into binary mark which was then embedded into quantified DCT coefficients in video encoding, then adopting watermarking technology based on side information, and lastly improving it according to the motion masking characteristic of human visual system, thus realizing the data hiding communication system based on platform OpenH323.