堆栈滤波的递归实现

堆栈滤波器是一类能够有效滤除脉冲类噪声，同时能较好地保护图象边缘与细节的非线笥滤波器，它所具有的 阈值分解和堆栈两个重要性质保证了滤波过程可以在二进制域并行实现，虽然ＭＡＥ（平均绝对误差）准则下最优堆栈滤波器可由ＬＭＡ（最小平均绝对误差）算法给出，但直接推导得到的滤波有属于非递归型滤波器，基于此，提出了ＭＡＥ准则下最优堆栈滤波器的递归实现方法，该方法不仅采取直接递归和最递归两种方案实现递滤波器的噪

Recursive Implementation of Stack Filtering

As a kind of nonlinear filters, stack filters can suppress impulse noise effectively, and keep edges and details at the same time. Those have two important properties: the threshold decomposition and the stack properties, which ensure that the filtering process go on in binary field parallelly. The optimal stack filters under mean absolute error (MAE) criterion can be developed with least mean absolute error (LMA) algorithm. The filters developed directly are non recursive filters. Based on the above, the recursive implementation approach of optimal stack filters under MAE criterion are advanced. Two approaches are used: direct recursive approach and optimal recursive approach, the latter resolve the matching between filter and input signals. The result of simulation with computer software illustrate that, the characteristics of noises removing of recursive stack filters are superior to that of non recursive filters, and the characteristics of edges and details preserving of the former are similar to the latter. So we can conclude that the advanced recursive approaches improve the function of the stack filters, which is meaningful to the advanced study for theory of stack filtering.