ABSTRACT
This chapter deals with order-based filters, which have been employed successfully to pass the desired signal (image) structures while suppressing noise. Complexity of the reduced solution depends on the model of the underlying signals, the nature of the additive noise to the signal, and the accuracy of the solution with respect to the above assumptions. The simplest and most important case of signal and noise is the additive white noise model where the signal s and noise v are assumed independent. When a single impulse exists in the signal, the mean filter spreads the impulse and reduces the amplitude. The rank-order filters have been used by statisticians for a long time and are simple modifications of the median filter. Gaussian and impulse noise are present, the trimmed mean filter becomes a compromise between mean and median filter.
