Homogeneous ANN using in image processing and picture generation

A systematic investigation of feasible dynamic structures in homogeneous artificial neural networks (ANNs) with lateral inhibition is presented. The main features may be studied by using a binary (Boolean) net consisting of bistable elements. A two-parameter coupling function describing lateral inhibition is used. It allows one to perform sufficiently complete, analytical ANN research and divide the parameter space into regions corresponding to image feature extraction and to picture generation as an example of a morphogenetic algorithm. It is shown how the procedure carried out in the human eye-brain system can also be used in artificial systems for image processing. The main abilities of such a net, namely the image key element extraction and content-rich spatial forms generation, are studied