Generalization of Order Statistic Filters: The Alpha-Trimmed Linear Filter

摘要

With the increase in use of digital communications has come a renewed interest in nonlinear filtering techniques. In some signal processing applications the suppression of noise on a discrete-time signal sequence can not be adequately accomplished using linear filtering. In order to preserve sharp edges in a signal while still smoothing noise and removing impulsive components requires some type of nonlinear or adaptive filtering. The most common nonlinear smoother to be investigated is the median filter. In this filter the median is taken from a finite length window of values surrounding the input value. This scheme has been shown to preserve edges in signals with a minimum of distortion and is resistant to outliers in the data. However, it often does not provide adequate smoothing of noise that is not by nature impulsive. Yet the desirable properties of median filtering make it worth investigating.