Conclusions - Logic-based Nonlinear Image Processing - page 144

hardware required. Examples of the implementation of stack filters, computational
morphology filters, and aperture filters have been presented.
The penultimate chapter covers the specific case of noise removal from astro-
nomical images, showing how the use of a training algorithm is essential to the cre-
ation of an appropriate nonlinear filter. With a manually-produced training set, it
was shown how an optimal soft morphological filter could be obtained. The chapter
also gives an example of real-time FPSA implementation of a spatio-temporal fil-
ter.

The objective of this topic was one of translation: from the language of mathe-
matics and set theory to the language of electronics and computer science. Many of
the powerful techniques outlined in this topic are not yet in common use for indus-
trial image processing. The author believes that this is because these techniques
have been developed primarily by mathematicians, and their descriptions reside in
texts that are not easily accessible to those who build industrial image processing
kits.
Ideally this topic has gone some way to correcting that situation. Some com-
plex areas have necessarily been glossed over, but if you wish to know more about
the complexities, there are plenty of good texts. A good starting point would be
Dougherty and Barrera's paper (referenced below) which bridges the gap between
pattern recognition theory and nonlinear signal processing. 1

I hope you have en-

joyed our journey, and thanks for reading to the very end.

Reference

1 E. R. Dougherty and J. Barrera, “Pattern recognition theory in nonlinear signal

processing,” Mathematical Imaging and Vision , 16 (3), 181-197 (2002).