On Large Local Error Accumulation in Multilevel Error Diffusion

Abstract

Error diffusion is an often used method that transforms a continuous tone (multibit) image into an image of lower bit depth, most commonly into a binary output of black and white. The simplicity of the processing and the quality of the output have made error diffusion a frequently used tool. Part of the image quality is attributed to the minimization of quantization errors in the error-diffusion process. This article describes local instabilities in a color multilevel error-diffusion system that—in contrast—can lead to large local errors in the output, far exceeding the normally expected quantization errors. This can have serious negative effects specifically in connection with the design and incorporation of color calibration sheets.