On generalization of order statistics based filters

Abstract

The generalization of order statistics (OS) based filters which are a useful class of nonlinear digital filters, is considered from two aspects. The piecewise-linear characteristics of OS based filters are first investigated, which indicates the direction of generalizing them in the set of piecewise-linear filters. The second way of the generalization considered here is through approximating an OS based filter by a filter of a general type. In both cases, the newly developed CPWL filter plays an important role.<>