一种盲图像分离算法

A blind images separation algorithm

提出了一种封闭式图像盲分离算法,用以解决当前盲图像处理问题.利用观测的混合图像二次特征函数二次导数矩阵与其对角分量和的二次导数矩阵联合近似对角化,估计未知的混合矩阵,进而用混合矩阵的逆阵与观测矩阵的乘积恢复原始图像,实现图像的盲分离.本算法无需数据存储和迭代,避免了采用高阶统计量的严重运算负担,分离效果显著,试验仿真结果验证了算法的有效性.

In this paper, a closedform images blind separation algorithm is proposed, to solve the problem of blind images separation. Through the observed mixing images, the unknown mixing matrix is estimated through approximate diagonalization by the second characteristic function second derivative matrix and second derivative matrix of its diagonal components summation. The original images can be recovered by the product of the observation images matrix and the inverse matrix of mixing matrix. BSS is reached. No data storage and iteration are needed in this algorithm, which avoids the heavy burden of computation through the high order statistics. The separation results are prominent. Simulation shows good performance.