The Relative Effects of Median and Mean Filters on Noisy Signals
E. Roy Davies · Journal of Modern Optics · 1992
This paper has studied the noise suppression qualities of median and mean filters, and has modelled the blurring effect of the mean filter. It has demonstrated that, when the signal-to-noise ratio is sufficiently low, the overall loss of accuracy introduced by the blurring effect of the mean filter can be more than cancelled out by the improved noise suppression properties it sometimes exhibits. However, this cannot occur for long-tailed noise distributions characterised by median relative noisiness index values of less than unity, for which the median filter is superior both in its noise suppression properties and also in its natural absence from signal distortion effects.