Digital implementation of cellular neural networks

Abstract

This paper presents a digital cellular neural network (CNN) for digital image processing applications. The CNN is a relatively new field in this research, making use of a high degree of parallelism to achieve higher levels of processing power which continuously paves new ways of how problems can be tackled. A digital architecture is employed due to the fact that digital devices allow for a very robust, yet simple and modular design while at the same time maintaining established performance standards. Digital design was carried out with VHDL using an iterative design methodology, meaning that only one out of several building blocks are chosen to ensure optimality, robustness and operational correctness. The main design objectives were to construct a digital CNN architecture which is fast and compact for digital image processing applications like next generation digital cameras.