基于邻域均值关系的Contourlet域量化水印算法

提出了一种基于邻域均值关系的Contourlet域量化水印算法。原始图像经过Contourlet变换，分解为一系列多尺度、局部性、多方向性的子带，对原始水印图像进行Arnold置乱加密，从分解的多方向高频子带中选择能量较大的子带作为嵌入区域，根据系数与邻域均值关系选择嵌入位置，用量化的方法修改嵌入位置的系数，达到嵌入水印的目的。水印的提取不需要原始图像，实现了盲提取。实验结果表明，该水印算法不仅具有较好的不可见性，而且对加噪、JPEG压缩、滤波、剪切等攻击具有较强的鲁棒性。

Quantization watermarking scheme based on neighboring mean value relation in Contourlet transform domain

This paper presents a Contourlet domain quantization watermarking scheme based on neighboring mean value ralation.After the Contourlet transform,the original image is decomposed into series of multi scale,local,and directional subim- ages,the original watermarking is scrambled by Arnold transform, and the larger energy high frequency is chosen as embedded domain.The embedded positions are chosen according to the mean value of neighboring symbol and coefficient relation. Then the embedded positions coefficients are modified by quantization scheme.The retrieving watermarking scheme is a blind detecting process,and it does not need original image.The experiment results show that this watermark scheme is good invisible and strongly robust to watermark attacks such as noise adding,JPEG compression,filter,and cropping.