本文在“压缩嵌入联合编码”框架下,提出一种基于秘密共享和信息分级保护的水印生成算法.与用算术编码压缩水印相比,本文的秘密共享方案在保证水印可认证性的前提下大幅度减少了信息嵌入量,同时对嵌入的信息依其重要程度进行分级保护,有效地增强了水印的鲁棒性.在此基础上,利用当前宏块与其周围宏块的运动矢量的相关性,提出一种基于运动矢量的自适应视频水印算法,可有效控制运动矢量的修改幅度,保证视频质量.实验表明该方案能保持很小的峰值信噪比损失和较好的视觉效果,且对于删帧攻击具有较强的鲁棒性.
Under the framework of joint information embedding and lossy compression, an algorithm based on secret sharing and unequal protection is proposed. Comparing with arithmetic coding which compressed watermark length, our algorithm can dra- matically reduce the amount of information bits embedded into video cover while guaranteeing watermarking authentication. Our al- gorithm not only improves compression efficiency but also adopts unequal protection strategy, thus improving watermark' s robust- ness considerably. Furthermore, by making full use of dependency among neighbouring macroblocks, an adaptive watermarking algo- rithm based on motion vectors is proposed to effectively control motion vector' s modification and guarantee video quality after wa- termarking. Experiments demonstrate that this algorithm not only keeps a small PSNR loss and satisfactory visual quality, but also shows resistance against frame-deletion attack.