摘要： 针对现有水印算法难以抵抗多种类型攻击的问题，提出一种基于非下采样Contourlet变换(NSCT)和奇异值分解(SVD)的多重数字水印算法。该算法采用不同的密钥对水印信息进行Arnold置乱，其宿主图像经过二层NSCT变换后得到大小相同的低频子带和高频子带。为提高算法鲁棒性，对各子带进行奇异值分解，将3个加密水印信息重复嵌入到具有最大奇异值的NSCT域低频和高频子带的子块中。在水印检测时，从3个提取结果中选取归一化均方误差最小的水印作为最终水印。实验结果表明，该算法能够有效抵抗JPEG压缩、椒盐噪声、剪切和中值滤波等多种类型的攻击，并且在保证水印不可见的前提下，提高了水印的鲁棒性和嵌入容量。

关键词: 数字水印, 非下采样Contourlet变换, 奇异值分解, Arnold置乱, 鲁棒性, 版权保护

Abstract: Aiming at the problem that those existing digital watermarking algorithms are difficult in resisting kinds of attacks, a multiple digital watermarking algorithm based on Non-subsampled Contourlet Transform(NSCT) and Singular Value Decomposition(SVD) is presented. The watermarking is scrambled by different keys, applying NSCT with two levels on the host image to get low and high frequency sub-bands with same size. It takes SVD operation for each sub-band, and embeds three encrypted copies of the watermarking into the maximum singular values of blocks in low and high frequency sub-bands of the NSCT domain respectively. In the extraction stage, the final watermarking is derived by choosing the minimum Normalized Mean Squared Error(NMSE) value amongst the three extracted watermarking. Experimental results show that the proposed algorithm can resist many types of attacks effectively, such as JPEG compression, pepper noise, cropping, median filter, etc, and it improves the robustness and embedding capacity of watermarking on the premise of guarantee the watermarking is invisible.

Key words: digital watermarking, Non-subsampled Contourlet Transform(NSCT), Singular Value Decomposition(SVD), Arnold scramb- ling, robustness, copyright protection

中图分类号: 

• TP309.2