Ancient Web: Øyvind Kolås's A Dither Preserved the Art of Tiny Pixel Patterns

Part of the Ancient Web Collection.

When a display has fewer available colors than a picture needs, programmers can hide the shortfall with carefully placed dots. Øyvind Kolås built a tiny way of deciding where the dots should go, then published the idea on a page that still invites readers to inspect the mathematics.

Visit the A Dither project.

Kolås describes A Dither as a small, spatially stable algorithm based on arithmetic and a few numerical constants rather than a large lookup table. He was interested in situations where an image must be reduced to a restricted palette or a display must update in small portions without unnecessarily changing surrounding pixels.

The page compares several techniques using an animated clip: traditional Bayer dithering, Kolås's addition- and XOR-based versions, and error diffusion. The examples reduce full-color video to the limited RGB 3:3:2 palette, with 256 possible colors. Another set shows how a grayscale ramp appears when only two levels are available.

A simple idea, honestly compared

The author does not claim that his result always looks best. In one passage he explains that the chosen clip deliberately exposes flaws in his method. In another he notes that alternatives based on prepared dithering masks can be faster or produce higher quality.

That willingness to publish trade-offs makes the page more valuable than a collection of before-and-after screenshots. Its compact code examples are accompanied by a comparison of uses: animated GIFs, electronic-paper displays, and software with very little space for graphics logic.

A surviving note says that as of March 25, 2014, FFmpeg's development version included the dithering modes. The page also displays an example FFmpeg command and links to a terminal image viewer that demonstrates related work. The prose is attributed to August 2013, fixing it as a specific stage in the development of these techniques.

This qualifies as a later specialist technical artifact rather than a 1990s homepage. Its value is that it preserves the original authorial explanation, sample images and limitations of a small algorithm that could otherwise appear only as a cryptic option in a much larger program.

Kolås expressly considers the algorithm public domain. That permission and the invitation to experiment with the constants fit an older spirit of web-based technical sharing: show how the method works and allow other people to improve it.

Compare the original animations and explanation of the algorithm.

Part of the Ancient Web Collection.