Input/output hardware strategies for cellular neural networks

In this paper hardware alternatives for the input/output circuits of a cellular neural network are discussed. Cellular neural networks are primarily applied in image processing applications. The realisation of a fully programmable 128/spl times/128 cellular neural network is possible in a standard state of the art CMOS process. Moreover, several alternatives for the integration of a solid-state image sensor that are compatible with a standard CMOS process have been published. The different possible input/output structures are evaluated towards their compatibility with both techniques. They provide the link towards the realisation of a smart image sensor and processor based on cellular neural networks.<<ETX>>