Abstract: Differing from wavelet transform，Contourlet transform has the ability of image multi-scale and direction decomposition using the anisotropic basis，thus it is better to capture the curve singularity and much more match the HVS properties.A unified robust and perceptual watermarking algorithm is presented with respect to two transform domain.The watermarks are adaptively embedded in each frequency sub-band，with the strength adjusted by the visual saliency of transform coefficients.Some experiments of such algorithm in two multi-resolution decomposition frameworks are given with respect to several common attacks.Experiment results show that the contourlet based algorithm has better safety，robustness and invisible compared with wavelet based algorithm.

摘要： 与小波变换不同，Contourlet变换通过各向异性基实现了图像的多尺度多方向性分解，更能捕获图像的曲线奇异性和匹配人类视觉特性。通过分解系数的视觉显著性决定水印强度，不同子带自适应地嵌入水印信号，提出了一种鲁棒性感知水印嵌入算法；该算法在小波分解和Contourlet分解的两种多分辨率框架下对常见水印攻击进行了实验测试。实验结果表明，在同等水印嵌入算法情况下，基于Contourlet和基于Wavelet变换的水印算法都能够抵抗常见攻击手段，但是Contourlet变换算法具有更高的安全性、鲁棒性和不可见性。