基于人类视觉感知的下采样转码视频水印*

提出了一种抵抗下采样转码攻击的压缩域视频水印算法，算法对下采样频域空间的对应中低频系数进行分析来确定服从正态分布的水印序列的嵌入位置，根据Watson视觉感知模型的计算来控制嵌入水印时系数的修改幅度，从而满足图像感知失真限制要求。实验结果表明，该水印算法对一般的噪声攻击和转码过程中最普遍的下采样处理具有较好的鲁棒性，水印的提取采取相关检测原理，不需要原始视频信息。

HVS-based video watermarking technique using downsampling

This paper presented a compressed domain video watermarking technique against video transcoding.The watermar-king approach was based on an downsampling algorithm working in the DCT-domain and the Watson visual model which introduced to control the watermark embedding strength. It chose the middle and low DCT coefficients as the position of watermark sequence to be embedded,in order to balance the relationship between watermark robustness and visual transparency.Experimental results show that the proposed strategies are useful in solving the transcoding attack,especially downsampling transco-ding.