Weighted Myriad Filters - Nonlinear Signal and Image Processing: Theory, Methods and Applications - page 176

(a) (b)

(c) (d)

FIGURE 5.10
(a) Noise-free image, (b) with 5% salt and pepper noise, (c) smoothed with 5 × 5 center weighted

median W c = 15, (d) smoothed with 5 × 5 center weighted myriad W c = 10 , 000, K = ( X

) + ( 21

X ( 5 ) )/ 2.

environments using FIR filter (smoother) design techniques, and then provide
the smoother with impulse resistance capabilities by means of myriadization.
The value to which K is to be reduced can be designed according to the im-
pulsiveness of the environment, for example, by means of an

- K curve.
It must be taken into account that a linear smoother has to be in “constrained
form” before myriadization can be applied. This means that the smoother
coefficients W i must be nonnegative and satisfy the normalization condition

α

i = 1 W i

1. A smoother for which i = 1 W i

= = 1 must be first “decomposed”