Title :
NFIR nonlinear filter
Author_Institution :
Dept. of Electr. & Comput. Eng., Shandong Polytech. Univ., Jinan, China
fDate :
7/1/1991 12:00:00 AM
Abstract :
A nonlinear filter is introduced whose output is nearest in distance to the output of a FIR (finite impulse response) linear phase filter; hence it is called the NFIR filter. After choosing the impulse response of the FIR filter properly, the NFIR filter can clean impulsive noise and preserve edges of signals. As a special case of the NFIR filter, a nearest mean (NEME) filter is derived. Theoretical analysis shows that the NEME filter is better than the median filter in smoothing additive Gaussian noise. In addition, based on the definition of the NEME filter, a new interpretation of the low-pass property of the median filter with window width 3 is described. Also, the performance of the nearest mean filter on a noisy edge is quantitatively compared with that of the median and mean filters
Keywords :
digital filters; filtering and prediction theory; random noise; FIR linear phase filter; NEME filter; NFIR nonlinear filter; additive Gaussian noise; finite impulse response; impulsive noise; low-pass property; median filter; nearest mean filter; noisy edge; signal edge preservation; signal processing; window width; Additive noise; Filtering algorithms; Finite impulse response filter; Gaussian noise; Image processing; Nonlinear filters; Signal analysis; Smoothing methods; Speech processing; Visual perception;
Journal_Title :
Signal Processing, IEEE Transactions on