Image Processing Reference
In-Depth Information
5
ALGORITHM OF BINARY IMAGE
PROCESSING
5.1 Introduction
In this chapter we present several algorithms for processing 3D images, in
particular for treating connected components and figures in a 3D image. Aims
of this chapter are:
(1) to understand processes essential in proceeding to the analysis of a 3D
figure in a 3D gray-tone image
(2) to learn the structure of algorithms through examples
In order to extract significant information from an input 3D gray-tone
image, we first need to segment a 3D figure that is likely to correspond to a 3D
object in the real world. Next, we extract features that have more condensed
information and derive from this a description of the figure. A binary image
is produced as a result of distinguishing figures from the background. Each
connected component corresponds to a 3D figure. If this segmentation of a
figure has been performed correctly, every connected component represents a
3D object meaningful in the real 3D world. This may not always be true in
practical image processing, but this is not currently important.
Information carried by a binary image is mainly in the shape of a figure.
Hence most algorithms presented here relate to processes based on the geo-
metrical properties of a 3D figure. Essential features of those algorithms are
transformations preserving connectivity or topology, extraction of the inside,
outside, and the border surface of a figure, and calculation of the distance on
a digitized image.
From the viewpoint of information concentration, we transform a massive
figure (a 3D figure) to a planar figure (the center plane and/or the border
surface), and transform a planar figure further to a linear figure (or a curve).
In another type of algorithm, we calculate the shortest distance from a voxel
inside of a figure to the background. We then concentrate the information to
local maximum points in the distribution of such distance values. Axis/surface
Search WWH ::




Custom Search