提出一种基于分块压缩感知和奇异值分解的双彩色图像水印算法。将彩色水印和载体图像分别进行RGB分离,采用分块压缩感知技术获得彩色水印在R、G、B这3个分量上的观测矩阵,实现对原始水印数据的压缩;将观测矩阵奇异值分解后的主成分以叠加扰动的方式调制在载体图像相应分量DWT-DCT混合变换后对角方向中频系数的奇异值上,DCT变换前采用Arnold变换置乱待修改的DWT中频系数,提高算法抗剪切性能。实验结果表明,该算法对噪声、中值滤波、JPEG压缩和亮度及对比度调整等攻击具有较强的鲁棒性,能抵抗一定程度的RST攻击。
A dual color image watermarking algorithm was proposed based on block compressed sensing and singular value composition.The color cover image and the watermark image were divided into three components R,G and B.To compress the original watermark image,the block compressed sensing sampling was performed with R,G,B components of the watermark image respectively to get the measurement matrix.Singular value composition was performed on the measurement matrix,and their principal components were added to the singular values obtained from the middle frequency coefficients after the correspon-ding R,G,B components of cover image had been performed with DWT-DCT hybrid transform.Arnold transform was performed on DWT middle frequency coefficients before DCT,which resisted cropping attacks.Experimental results show that this algorithm is robust to some common attack operations such as adding noise,JPEG compression,brightness-and-contrast adj usting and so on.