Realisation of a digital cellular neural network for image processing

A a digital cellular neural network (DCNN) based on the SIMD-architecture is presented. The network is optimized for image processing applications. Due to the massive parallel architecture of the global structure and due to the local parallel operating blocks of the cells, high calculating speed can be obtained. Processing of images with sizes up to 100/spl times/100 pixels in realtime is principally possible. In order to process large images, which are much greater than the physical network, virtual processing is needed, and supported by the hardware. As prototype, a cascadable net of 2/spl times/2 cells is implemented on a chip using the 1.0 /spl mu/ process of ES2.<
>