Video Watermarking Based on Motion Vector Statistic
Zhuang Qian-yu · Harbin Ligong Daxue xuebao · 2010
Temporal synchronization is one of the challenges for blind watermark detection.If the detector's input is watermarked but synchronization fails,the embedded watermark will not be detected.For attacking temporal synchronization efficiently,a novel Mpeg4 video watermarking algorithm is proposed based on data statistical classification of motion vector and Mpeg4 compressed format.According to motion vectors'statistical characteristics,we choose motion vectors to embed watermark.The algorithm avoids the sensitivity to watermark synchronization compared with traditional algorithm.The experimental results demonstrate the algorithm has great robustness and perceptual transparency.