基于压缩感知的图像水印算法研究

提出一种基于压缩感知和非采样Contourlet变换的数字图像水印算法.首先使用NSCT变换对载体图像进行稀疏分解,对低频系数矩阵进行Arnold置乱;然后采用高斯观测矩阵对系数矩阵进行压缩感知测量,在感知域(压缩信号)中进行水印嵌入;最后使用OMP算法恢复重组的低频矩阵信号,使用NSCT逆变换重构图像.算法中,测量矩阵同时充当了密钥的作用,增强了算法的安全性,此外Arnold变换也增强了水印的不可见性和安全性.仿真实验证明,所提算法对加噪、滤波、旋转具有较高的鲁棒性.

Research on an image watermarking algorithm based on compressed sensing

This paper proposes a new algorithm based on Non-sampling Contourlet Transform(NSCT) and Compressed Sensing(CS) to improve the watermarking's attack robustness.First,it uses NSCT to transform the image and scramble the low-frequency coefficients by Arnold scrambling,then it uses the compressed sensing for the low-frequency matrix of scrambling the watermark embedded in the compressed signal;it finally recovers the reorganization of the low-frequency matrix signal by OMP algorithm,and uses NSCT inverse transformation to reconstruct it.Experimental results show that the new algorithm has a high robustness in noise,filtering and rotation.