Page 198 - Intelligent Communication Systems

CHAPTER 12 /COMPUTERVISION 169

FIGURE 12.15 Coding

(2) When a node with no leaf is reached, the value "1" is output and the
process proceeds to the next step.
(3) When a leaf is reached, the value "0" and a color code (black or white)
are output and the process goes on to the next step. When a leaf is a
pixel, only a color code is output.
In Figure 12.15, the following coded information is created:
10WOB1WBWWOW

12.4.3 Image Recognition
Computer vision provides the functions for recognizing and identifying an image
as a specific object, such as a house, a human, or a road. Human beings recognize
an object using their knowledge of the object (e.g., a house, a human, or a road).
It is necessary to develop image-recognition technology that can recognize all
kinds of objects.

12.4.3.1 Course of Pattern Recognition
The course of pattern recognition is shown in Figure 12.16. After input of an image,
noise is eliminated by smoothing, and the size of the image is normalized by using
a pyramid data structure. The features of the image are obtained by means of
filtering or Fourier transformation. Apattern that matches a standard form is chosen.
The most suitable pattern is selected by using dynamic programming and the dis-
tance between the input image and the standard image.

12.4.3.2 Pattern Matching
Calculating the Distance Between an Input Image and a Standard Image: We
calculate the distances between input image feature points and those of a
standard image, or the distances between the input-image's Fourier
transform coefficients and those of the standard image. For example, given