Median filters

Abstract

Most modern digital filtering is done by taking the average (mean) of a signal or some weighted average. Another method is to use feedback, which more closely resembles how analog filters with feedback operate. In the case of low pass filters, all these methods tend to give a trade off in getting the signal to pass while attenuating the higher frequency noise. An alternative is to use a median filter, which selects the mid value of a group of points. While this is not as computationally simple as other filters, it allows for the attenuation of noise while allowing sudden changes in signal level to pass thru unaltered. This paper discusses the characteristics of median filters and methods of implementing them. © International Foundation for Telemetering, 2006.