一种新颖的基于非压缩数字视频的水印盲检测算法

数字水印是一种嵌入到多媒体数据中用来进行版权标识的工具。在众多的宿主媒体中，数字视频因具有隐藏容量大、透明性好、鲁棒性强等诸多优点而受到日益广泛的关注。但是很多文献中提到的视频水印都是从数据流中提取单帧图像进行处理，这类算法与静态图像的水印方法如出一辙，没有充分利用视频文件的各种特性。而且对帧平均、视频压缩等常见的运动图像攻击方法十分敏感。针对这些问题，本文以非压缩视频文件为实验对象，结合人类视觉模型和彩色图像场景分割的方法，提出并实现了一种基于视频时间轴的数字水印盲检测算法。实验结果表明算法有效实用。

A Novel Watermark Blind Detection Algorithm for Non-Compressed Digital Video Signal

Digital watermark is a powerful tool for copyright protection of multimedia products. Nowadays, research efforts have increasingly been focused on watermarking in digital video signal due to its large watermark hiding capacity and the watermark embedded having many excellent features such as imperceptibility and strong robustness. Video watermarking in most literature is handled frame by frame individually. In principle, these methods are not different from still image watermarking for temporal redundancy of video data has not been fully utilized. Therefore, watermarks embedded by those methods are very sensitive to some common attack schemes, such as frame averaging and video compressing. To resolve these problems, a novel watermark oblivious detection algorithm is proposed in this article. According to this algorithm, theories of Human Visual System (HVS) and Scene Segmentation are considered. Different from conventional video watermarking schemes, the proposed algorithm is implemented in temporal domain. Simulation results indicate that the proposed algorithm is effective and robust .