Recursion adaptive order statistics filters for signal restoration

A recursion-adaptive nonlinear filter structure that is based on the linear combination of order statistics is developed, using the minimal mean square error (MSE) criterion and the unbiasedness condition, Two efficient methods to update the filter coefficients are presented. The proposed recursion algorithms can adapt well to a variety of noise probability distributions and are useful for real-time applications. The MSE estimation computed using the local statistics of the input signal and noise is considered. Simulation results on one-dimensional signals and real images are also presented.<>