Error Diffusion on an Adaptive Raster

Whenever a continuous-tone image has to be displayed on a bi-level medium, some kind of halftoning has to be carried out. Nowadays processing of the continuous-tone image is often done by a computer, i.e. the sampled and quantized continuous-tone image is transformed into a binary image by a halftoning algorithm. In the following we will assume the continuous-tone image to be scaled between 0 and 1. In addition, we will restrict the discussion to algorithms for output devices where the addressability of the pulse location is approximately equal to the pulse extension. In this case the binary picture can be described by a pattern of independent black and white pixels. Depending on the properties and needs of the output device, pulse-width modulation (PWM) or pulse-density modulation (PDM) is used. A special kind of PDM is generated by the error diffusion algorithm (ED) [1]. Because of its easy implementation and its relatively high speed, this algorithm is in widespread use.