Abstract

Previously, we have presented a new method of optical analog-to-digital conversion based on oversampling and error diffusion coding techniques and demonstrated a firstorder error diffusion modulator using multiple quantum well modulators.¹ In this paper, we extend this idea to multidimensional error diffusion coding and, specifically, optical digital image halftoning. Although error diffusion coding can be shown to provide an optimum solution to this problem subject to a fidelity criterion, electronic realizations suffer from implementation constraints. In an electronic realization, the algorithm raster scans the image, and for each pixel, a binary quantization decision is made based on the intensity of the individual pixel and the weighted error from pixels in the diffusion region. As a result, the diffusion filter is necessarily causal resulting in undesirable visual artifacts, and substantial memory storage is required. By using an optical implementation of this algorithm, each pixel quantization decision can be computed in parallel and therefore the diffusion filter need not be causal. Visual artifacts resulting from the causality of the diffusion filter can therefore be reduced. Also, the inherent parallelism associated with optical processing can reduce the computational requirements while decreasing the total convergence time of the halftoning process.

© 1993 Optical Society of America