一种多级离散余弦变换和奇异值分解的零水印算法

摘要

传统的基于离散余弦变换的数字水印方法在嵌入水印时都要对原始图像数据进行一定的修改导致嵌入水印图像的保真度下降.文章提出了一种基于多级离散余弦变换和奇异值分解的零水印算法,该算法首先对要嵌入的原始图像进行多级离散余弦变换(MDCT),提取低频子图并对其进行奇异值分解(SVD);其次对嵌入的水印图像进行置乱,将置乱后的图像进行奇异值分解;最后将原始图像和水印图像两次SVD后的结果进行运算产生密钥图像.由于本文选取的水印为灰度图像而不是伪随机序列或二值图像,因此在水印检测时具有很好的区分度和检测精度.仿真实验表明该算法的鲁棒性优于现有算法,而且在抗各种常见攻击方面性能良好.%The drawback of the conventional digital watermark methods based on multiple level discrete cosine transform is loss of image fidelity due to the fact that essential modification to the original image should he done in embedding digital watermarks. A new algorithm for zero digital watermarks based on multiple level discrete cosine(MDCT) and SVD transforms is proposed in this paper. Firstly extracting the low frequency part of the multiple level discrete cosine transform result of the original image and SVD is applied. Then the watermark image is confused with the image scrambling,meanwhile SVD is applied as well as. Finally the encrypted image is constructed by correlative operation for the SVD results of original image and watermark image. The proposed algorithm is relatively robust in the presence of noise and presents high detection accuracy due to the fact that the watermark is a gray level image instead of a sequence of binary number. Experimental results show the superiority of the proposed method in comparison with some recent methods. It is also shown that the proposed method is relatively robust to some image degradation processes such as noise, cropping, filtering and JPEG compression etc.