彩色图像四元数频域奇异值分解水印算法

Color Images Watermarking Algorithm Based on Quaternion Frequency Singular Value Decomposition

将四元数傅里叶变换与四元数奇异值分解技术相结合并引入到对彩色图像的水印处理,提出一种基于四元数频域奇异值分解的彩色图像盲水印算法.首先对彩色载体图像进行分块并采用四元数傅里叶变换(quaternionFourier transform,QFT)得到其频域矩阵,然后对频域矩阵中的单位小块进行四元数奇异值分解(quaternion singularvalue decomposition,QSVD),得到实系数奇异值,使用奇偶量化调制法将水印信号嵌入到单位小块的最大奇异值中.仿真实验结果表明,嵌入的水印分布在空域图像各彩色分量中,在不可见性以及鲁棒性的比较中优于传统的彩色图像亮度域以及独立多通道处理方法.

By combining quaternion Fourier transformation with quaternion singular value decomposition and introducing them into the color image watermarking processing, a blind color image watermarking algorithm is proposed based on the quaternion singular value decomposition in frequency domain. The color host image is divided into blocks and the relative matrix in the frequency domain is obtained by the quaternion Fourier transformation firstly. Then the unit blocks of the matrix in the frequency domain are decomposed by quaternion singular value decomposition （QSVD） to get singular values with real coefficients. The method of odd and even quantization index modulation is applied to embedding the watermark signals into the maximal singular value of each unit block. Experimental results demonstrate that the watermarks are spread into all color components in the spatial domain and the proposed watermarking scheme is better than traditional luminance watermarking algorithms and separate channel watermarking algorithms in robustness and imperceptibility.