M. Muneyasu et al., A 2-D IIR NEURAL HYBRID FILTER FOR IMAGE-PROCESSING, Electronics and communications in Japan. Part 3, Fundamental electronic science, 77(2), 1994, pp. 75-84
A two-dimensional (2-D) IIR neural hybrid (2DINH) filter is proposed f
or image processing. This 2DINH filter consists of the cascade connect
ion of a 2-D IIR linear filter followed by a neural filter. The propos
ed filter can be applied to smoothing for the images corrupted with Ga
ussian noise and/or impulsive noise and to edge detection for the imag
es corrupted with Gaussian noise. The parameters of a neural filter ar
e adjustable by a learning algorithm to adapt itself to the property o
f an image to be processed. Finally, some examples are given to illust
rate the utility of the proposed filter.