Color halftone error-diffusion with local brightness variation reduction

Abstract

Error diffusion algorithms such as the celebrated Floyd Steinberg error-diffusion algorithm are high-performance halftoning methods in which quantization errors are diffused to 'future' pixels. Originally intended for grayscale images, they are traditionally extended to color images by error-diffusing each of the three color planes independently (separable error-diffusion). Adding a design rule which is based on certain characteristics of human color perception to the error-diffusion paradigm results in a color halftoning algorithm having output of considerably higher quality when compared to separable error-diffusion. These benefits are achieved by adding the Minimum Brightness Variation Criterion (MBVC) to the design rules of color error-diffusion halftoning methods. Halftone values are constrained to be vertices of a Minimum Brightness Variation Quadruple (MBVQ) associated with each pixel of the color image being processed. The algorithm presented requires no additional memory and entails a reasonable increase in run-time.