基于变步长的混合图像自适应盲分离算法

Adaptive Blind Separation Algorithm of Mixed Image Based on Variable Step-Size

根据混合信号的分离程度,提出了一种新的图像信号分离算法,通过建立分离矩阵控制步长因子变化,利用自适应不完整自然梯度法实现图像的有效分离。新算法既很好地解决了固定步长算法收敛速度和稳态误差之间的矛盾,也解决了其他变步长算法须选择较小初始步长才能实现分离的问题。仿真结果表明新算法收敛速度快,稳态误差小,综合分离性能明显优于其他算法。

According to the separation level of mixed signal,a new separation algorithm of image signal is presented.The separation of images can be effectively achieved through the establishment of controlling step factor change in the separation matrix and the adoption of adaptive nonholonomic natural gradient method.The new algorithm serves as an excellent solution to solve the contradictions between the convergence speed and the stability error of the fixed step-size algorithm.It also settles the seperation problem without choosing a smaller initial step size.The simulation results demonstrate that the convergence speed and the steady-state errors of the new algorithm are improved and its comprehensive separation performance is obviously superior to those of other algorithms.