Some separable linear filtering tasks using CNNs

Abstract

In this paper, we propose some efficient realizations of separable 2-D spatial filters implemented on Cellular Neural Networks (CNNs), based on the Gaussian distribution function, which is approximated by both FIR and IIR filters. We also present a method of iterative filtering, which allows a selective Gaussian function to be implemented by repeating a simple filtering task several times. Some examples of selective low-pass and band-pass separable filters are given to illustrate the design methods.