一种基于相关值检测盲水印方法  

A BLIND WATERMARK SCHEME BASED ON SVD

作　　者：董兵[1] 张建伟[1] 

机构地区：[1]南京信息工程大学数学系,江苏南京210044

出　　处：《计算机应用与软件》2007年第4期162-164,共3页Computer Applications and Software

摘　　要：针对传统的相关值检测盲水印算法嵌入信息量小,提取水印正确率低的缺点,提出了一种结合分块SVD分解的相关值检测盲水印方法,它在嵌入过程中对原始图像进行分块SVD分解,取最大的一部分奇异值重新构成载体图像块,将水印置乱后,水印信息通过向载体图像块叠加一个高斯分布的模板实现嵌入,在保证透明度的同时增大了嵌入量。为提高水印提取的准确率,首先通过对待检测图像滤波预处理,然后通过比较相关值的大小来判断水印信息。实验结果表明,新的算法在保证水印的不可见性的同时,增加了嵌入信息量,提高了检测的准确性,且对于常见的图像压缩、噪声和剪切等攻击具有良好的鲁棒性。The original blind watermark based on correlativity has the disadvantage of small capacity and detection false. In this paper,we improve a blind watermark algorithm based on SVD. First ,we made the original cover data singular value decomposition,use some big singular values and respective singular vectors to rebuild cover data. Then, watermark is embedded by adding a Gaussian distribution pattern on cover data. For Imperceptibility and bigger capacity, we split the original image into blocks and use one Gaussian pattern in embedding. For detection accurate, we use a filter on the image before detected. Compare the correlation of every block instead of use a threshold value. The experimental results show our algorithm has very excellent Imperceptibility, and robust to some popular attacks like JPEG compression,image noise and image cropping. The original image was not required for extracting the embedded marks.

关 键 词：数字水印 奇异值分解 盲水印 版权保护 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]