基于SVD和ICA的鲁棒水印算法

A Robust Watermarking Scheme Based on SVD and ICA

提出了一种新的基于SVD(singularvaluedecomposition,奇异值分解)和ICA(independentcomponentanalysis,独立分量分析)的鲁棒水印算法。采用二值图像作为水印,并进行Arnold变换,提高了安全性。SVD用于水印的嵌入过程,ICA用于水印信号的提取。该算法的水印容量大大提高,同时还保证了鲁棒水印的不可感知性。借助于ICA,该算法在不需考虑水印图像所经历的攻击类型及攻击参数的情况下,能够正确提取水印信号。实验结果表明,该算法对于JPEG压缩、噪声等具有鲁棒性,尤其对于几何攻击如旋转、剪切、伸缩具有很好的鲁棒性。

A novel watermarking algorithm for the digital image is proposed. The Singular Value Decomposition (SVD) is employed during the embedding process of the watermarking. Instead of random sequences, some readable signature is used as watermark. By means of Independent Component Analysis (ICA), the watermark is successfully extracted. In the proposed algorithm the capacity of the embedded watermark is enhanced, and at the same time the transparency can be guaranteed. Our method can extract the watermark without taking attacks into consideration. Experiment results have demonstrated that the proposed approach is robust against the common signal processing, such as JPEG compression, additive noise. In particular, it is also robust against geometric attacks, such as rotation, cropping, scaling etc.