Pointillist Halftoning

We explore an extension of current halftoning algorithms that aims to reproduce pointillism, a style of postimpressionist painting. Halftoning algorithms solve the problem of reproducing an image with many colors on a device with very few colors (usually only two, black and white). Pointillist painters also created images using a very limited number of paint colors. Both halftoning and pointillism use dot patterns to suggest intermediate colors not available on the image created. In order to reproduce pointillism, we extend error diffusion, which is a commonly used halftoning algorithm. Error diffusion relies on an ordered grid of pixels, which appear as dots of color on the final image. If these dots are initially not on a grid, but are placed randomly, a new algorithm is needed. Our solution is to build a graph of pixels with links between neighbors on the image, and replace row-by-row processing with graph traversal. This extension of error diffusion is the main contribution of the paper. Beyond simulating an artist’s technique, our extension improves color reproduction, by producing alias-free dot patterns and more highly saturated colors for offset color printing.