Nonlinear Image Processing

Nonlinear Mean Filters and

Their Applications in

Image Filtering and Edge

Detection

CONSTANTINE KOTROPOULOS, MICHAEL PAPPAS, AND IOANNIS PITAS

Department of Informatics

Aristotle University of Thessaloniki

Thessaloniki, Greece

5.1 Introduction

Noise filtering and edge detection are two of the most important tasks in image

processing. Noise removal in images is a particularly difficult task due to the

nonstationary nature of images and the different types of noises (e.g., additive

and signal-dependent noise) that corrupt images.

In this chapter, we study the class of nonlinear mean filters for noise removal

and edge detection. Nonlinear mean filters can be considered as an alternative

to the median filter and its extensions and/or generalizations (e.g., max/median,

multistage median, median hybrid, trimmed mean, L-filters, etc.) [Ast97, Pit90].

All of these filters offer improvements in certain statistical properties at the ex-

pense of a lower ability to remove impulse noise.

135