Application of Recurrent Neural Network for Generating Grayscale Digital Half-Tone Images

Abstract

Clustered dot ordered dithering (CDOD) is one paradigm of digital half-toning that is employed for systems which cannot produce more than two levels at display. The conventional CDOD results in major limitations i.e. visually objectionable periodic patterns, false contouring and blurred appearance of the half-tone images. Cluster generation using artificial intelligence may be a potential solution. In this paper recurrent neural network (RNN) based framework for adaptive cluster generation has been proposed. Under RNN, Elman model (Elman RNN) and Jordan model (Jordan RNN) have been employed. The implementation steps of the proposed algorithm, along with the results, have been presented. The results show that this method can avoid the major limitations of conventional CDOD methods, particularly appearance of periodic patterns and may be potentially useful in the field of digital half-toning.