Image Processing Using Pulse-Coupled Neural Networks

Abstract

VIIIPreface to the First Edition out. In other words, each neuron was connected to several other neurons. In electronics a different "wire" is needed to make each connection and large networks are quite difficult to build. The PCNN, on the other hand, has only local connections and in most cases these are always positive. This is quite plausible for electronic implementation.The PCNN is quite powerful and we are just in the beginning to explore the possibilities. This text will review the theory and then explore its known image processing applications: segmentation, edge extraction, texture extraction, object identification, object isolation, motion processing, foveation, noise suppression and image fusion. This text will also introduce arguments to its ability to process logical arguments and its use as a synergetic computer. Hardware realisation of the PCNN will also be presented.This text is intended for the individual who is familiar with image processing terms and has a basic understanding of previous image processing techniques. It does not require the reader to have an extensive background in these areas. Furthermore, the PCNN is not extremely complicated mathematically so it does not require extensive mathematical skills. However, the text will use Fourier image processing techniques and a working understanding of this field will be helpful in some areas.The PCNN is fundamentally unique from many of the standard techniques being used today. Many techniques have the same basic mathematical foundation and the PCNN deviates from this path. It is an exciting field that shows tremendous promise.