Research On Robust Digital Video Watermarking

With the development of Internet,it's more and more convenient to watch and transmit video content,which has become the essential part in people's work and daily life.Meanwhile,the widespread of pirate video is damaging the economic interest of video creator and their passionate,causing copyright disputes.Therefore,how to determine ownership of video is an important research subject.Digital video watermark technology is to embed copyright related information into video content or to find a characteristic of video and keep it recorded,then copyright information can be extracted and identified after video is pirated,proving the ownership of video and protecting copyright.This work focus on digital video watermark algorithm.We first introduce the concept of digital video watermark,then summary some attack on video watermark,e.g.frame insertion,frame deletion,bitrate change,etc.Then We propose two robust digital video watermark algorithms which can resist those attacks.We propose a video zero watermark algorithm which resists geometric attack.The algorithm partitions frame sequence into groups,then performs discrete cosine transform in the time domain,takes the first coefficient?low frequency component?to construct characteristic matrix,counts ₁n,the number of elements which is greater than8 and ₀n,the number of elements which is lower than-8,characteristic of one group is determined by polarity of ₁n and ₀n,characteristic of video is the concatenation of group characteristic.Experiment results show that this algorithm is very robust against video rotate,crop,scale attack,it's also robust against noise and average blur attack.The second algorithm is based on blocked DCT.Each video frame is partitioned into 8?8 blocks,then DCT is performed on those blocks,coefficients in each block are modified based on one watermark information bit so that a certain size relation between two coefficients in each block holds.To increase the robustness of algorithm,one information bit is embedded into one piece of video that lasts time t.In the extraction phase,it counts the number of blocks which the size relation holds and determines which bit is embed into this frame using majority decision,obtaining a bit string,then decodes bit string and detects watermark information in it.Experiment results show that this algorithm is robust against frame copy,frame deletion,bitrate change,frame rate change,etc.It also has a good invisibility.