Information Technology Reference
In-Depth Information
sampling and digitizing them. In essence, this is the analog form of information
processing in which the capability of information processing by the human cerebral
cortex is simulated.
It should be noted that the main methodological difficulty in using reaction-
diffusion devices lies in the fact that the understanding of the mechanisms used by
the brain for solving specific problems leaves a great deal from being desired. If
they were known, modeling these mechanisms by the media with an architecture
similar to the Grossberg neural networks would allow to create effective devices
functionally similar to the cerebral cortex. Our ignorance of these mechanisms,
combined with the von Neumann principles of information processing entrenched
in the minds, often leads to inefficient algorithms. Therefore, devices based on them
do not exploit the basic fundamental advantages of reaction-diffusion media. At the
same time, a correctly guessed algorithm for information processing by a reaction-
diffusion medium leads, on the one hand, to an effective device and, at the same
time, hopefully sheds light on the mechanisms of functioning of the cerebral cortex.
In 1986 the German researcher Lothar Kuhnert discovered a remarkable prop-
erty of light-sensitive Belousov-Zhabotinsky media. For a sufficiently long time
(minutes) they retained the image that appears in the medium when this image is
projected onto its surface. It was also found that during its existence in the medium,
the picture entered undergoes evolution, in the course of which its contour is
periodically detected and transitions between the negative and the positive forms
of the image take place. This effect was described in an article published by
L. Kuhnert with the Pushchino Research Center researchers K. I. Agladze and
V. I. Krinsky, which the authors called “Image processing using light-sensitive
chemical waves.” Later, physical and informational aspects of this problem were
studied in detail by the International Institute of Control Sciences and the physical
faculty of the Moscow State University.
These studies have shown that, indeed, in many cases, the results of the image
evolution in the medium are equivalent to one of the image processing operations
used in practice. Nevertheless, several issues remained open which play an impor-
tant role in the practical application of chemical reaction-diffusion media for image
processing:
What is the basis of the similarity between the results of image evolution in the
medium and the operation of its processing?
Does this evolution always select only the graphic elements of the image
previously entered into the medium, or can features not present in the original
image emerge in the course of its evolution?
Answering these questions was made possible through a detailed comparative
analysis of image processing by chemical reaction-diffusion media and the method
of mathematical morphology.
Let us consider the results of this analysis, starting with the processing of black-
and-white images by the Belousov-Zhabotinsky medium, whose composition
corresponds to its functioning in an excitable regime.
