摘要
针对视频水印常见的共谋攻击和时间同步攻击两种攻击方式,在对逐帧嵌入方法抗时间同步攻击问题的基础上提出了一种基于镜头分割的视频水印模型;通过在各帧图像变化不大的场景内嵌入高度相似的水印信息,在各场景之间嵌入不相关的水印信息达到抵抗共谋攻击的目的;通过有限状态机来生成用于各个镜头的嵌入密钥,在检测时有限搜索这些密钥实现时间同步。仿真结果表明,该算法不仅具备了较好的抵抗共谋攻击的能力,同时还提高了逐帧嵌入方法抗时间同步攻击问题的算法对抗帧删除攻击的能力。
A frame-to-frame video watermarking scheme is presented based on scene-division, which has the capability to resist any normal attack, especially the collusion attack in this paper. In the proposed scheme, the collusion attack is resisted mainly through embedding watermarks, which have high-level similarity to approximately invariant scenes and watermarks being uncorrelated to different scenes. Meanwhile, the embedded key is generated for each scene by limited state machine, and can be used to achieve time-synchronization by searching them in watermark detection.
出处
《西华大学学报（自然科学版）》
CAS
2007年第6期38-41,共4页
Journal of Xihua University:Natural Science Edition
关键词
视频水印
共谋攻击
时间同步
video watermarking
collusion attack
time-synchronization
