Multilayer transient-mode CNN for solving optimization problems

An implementation of analog parallel network for solving image optimization problems is presented. The implementation is based on a multilayer cellular neural network (CNN). A general optimization procedure is divided into subfunctions each of which is realized by a network layer. The problem of multilevel halftoning of images is used as an example of the optimization procedure. The resulting network which consists of four layers is described in the paper. The network was simulated on a digital computer and its performance was evaluated using a set of test images. Results of the tests and other potential applications of the proposed network are discussed.