The Cellular Neutral Network (CNN) paradigm is a framework of locally connected analog, nonlinear, dynamic processing elements. CNNs are useful in many application areas including image processing, robot control, biological modelling, etc. An optical implementation of the CNN model using 'liquid crystal light valves' (LCLV) has some definite advantages (especially higher speed and higher resolution) over a VLSI implementation. However the exceptional computing speed (the speed of light) can be fully exploited when 'feedforward-only' templates alone are used. In this work we show how the processing capabilities of optically-realized feedforward-only CNNs can be increased by combining several serial or parallel operations into a single feedforward template. We also derive conditions on a feedforward template whereby the processing resolution in an optical implementation can be increased.
|Number of pages||9|
|Journal||AEU. Archiv fur Elektronik und Ubertragungstechnik|
|Publication status||Published - May 1 1992|
ASJC Scopus subject areas
- Computer Networks and Communications
- Electrical and Electronic Engineering