基于压缩感知观测值的数字图像水印算法

根据压缩感知理论具有计算保密性这一特点,提出一种新的基于压缩感知观测值的数字图像水印算法.首先对载体图像进行小波变换,得到稀疏后的小波系数矩阵;然后对小波系数矩阵的不同频率部分,用不同的观测矩阵进行压缩感知,得到压缩后的观测值;再将水印嵌入至小波高频系数部分的观测值中,使用子空间追踪算法恢复稀疏信号,进而通过小波反变换得到加密图像.实验结果表明:该算法具有信息安全性,能满足水印不可见性和鲁棒性要求;相比同类算法,该算法的水印提取过程更加灵活与安全.

Watermarking algorithm for digital image based on compressive sensing measurements

According to the computational secrecy property of compressive sensing(CS) theory,a new digital image watermarking algorithm based on CS measurements was proposed.First,the wavelet transform had been made to the original image,and got the wavelet coefficients matrix.Second,according to the different frequency the coefficients represented,each part of the matrix was observed by a different measurement matrix to obtain the measurements.Then the watermark was embedded into the high-frequency-parts measurements.At last,a subspace pursuit(SP) algorithm was used to restore the sparse signal and yet make the inverse wavelet transform to get the encrypted image.The experimental results indicated that the algorithm possessed the information security,and meet the need of the transparency and robustness.By compared with other similar CS-theory-based digital watermarking algorithms,the watermarking extraction process was more flexible and safer.