Analysis and VLSI architecture of a nonlinear edge-preserving noise-smoothing image filter

Abstract

A nonlinear, nonrecursive, two-dimensional filter for edge preserving and noise smoothing of images is developed. The filter is applied over a 3 × 3 window of pixels and uses a thresholding operation. It is tested against other filters of the same class and its efficiency is proven by simulation. Analytical evaluation is also performed using a first-order stochastic Markov model. The computational requirements lead to the use of special purpose hardware for real-time implementation. A massively parallel array structure is proposed and the basic processing element realising the filtering operation is implemented in CMOS VLSI using minimal hardware components.