TABLE OF CONTENTS In this chapter we will discuss neighborhood operations for performing the elementary task of averaging. This operation is of central importance for low-level image processing. It is one of the building blocks for the more complex feature extraction operators discussed in Chapters 13 15.
In the simplest case, objects are identified as regions of constant radiance, i. e., gray values. Then, averaging gives adequate mean values of the gray values within the object. This approach, of course, implies a simple model of the image content. The objects of interest must indeed be characterized by constant gray values that are clearly different from the background and/or other objects.
However, this assumption is seldom met in real-world applications. The intensities will generally show some variations. These variations may be an inherent feature of the object or could be caused by the image formation process. Typical cases are noise, a non-uniform illumination, or inhomogeneous background.
In complex cases, it is not possible to distinguish objects from the background with just one feature. Then it may be a valid approach to compute more than one feature image from one and the same image. This results in a multicomponent or vectorial feature image.
The same situation arises when more than one image is taken from a scene as with color images or any type of multispectral image. Therefore,...
