Cellular neural network for automatic multilevel halftoning of digital images

An implementation of a fully automated multilevel halftoning algorithm using cellular neural network (CNN) is presented. The algorithm tracks the transient output of CNN limited to a small number of grey levels and selects the image that has the best visual appearance using the model of the human visual system (HVS) and the mean square error criterion. The algorithm is implemented in the form of a three-layer CNN. The first layer performs halftoning optimisation of the input image. The second layer approximates the HVS filtering. The third layer selects the best multilevel halftoned image during the transient of the first layer. In addition, the third layer has an associated logic that stops the transient of the first layer when the optimum image is achieved. Results of the software implementation of the proposed algorithm are presented.