dither

Dithering is a technique for representing colors or tones that a display or printer cannot physically produce. By laying down a precise pattern of dots in colors that the system can produce, it tricks the human eye into blending those dots together and perceiving an intermediate shade. This matters most in constrained environments: early computer displays limited to 256 colors, thermal printers that can only mark or not mark paper, or industrial control panels with monochrome LED arrays.

The most common dithering algorithms scatter the quantization error from one pixel across neighboring unprocessed pixels, rather than simply rounding each pixel to the nearest available color. Floyd-Steinberg dithering, developed in 1976, distributes error in a weighted pattern to the right and below the current pixel. Ordered dithering uses a fixed threshold matrix (Bayer matrix) applied uniformly across the image; it is faster and more predictable but often produces visible checkerboard artifacts. Threshold dithering, the simplest form, applies a single cutoff: if a pixel value exceeds the threshold, mark it; otherwise leave it blank.

Where it matters in industrial work

Industrial applications rely on dithering wherever output is binary or severely limited. Thermal label printers use dithering to render photographic images as patterns of printed and unprinted dots. Fax machines historically applied dithering to convert grayscale camera input into black-and-white line images for transmission. Dot-matrix printers used dithering to create the illusion of continuous tone across multiple passes with different pin patterns. Industrial vision systems and optical inspection cameras often use dithering when converting continuous analog sensor data into discrete digital commands or display states.

The cost of dithering is complexity and processing time. A raw raster scan simply rounds each pixel value to the nearest available output; dithering requires additional memory to hold error buffers and extra arithmetic per pixel. In real-time systems running on embedded hardware, ordered dithering wins because lookup tables replace computation. Conversely, offline image preparation, such as preparing artwork for silk-screen masks or thermal transfer plates, can afford the cost of better algorithms.

The term originates from the musical meaning: a slight, rapid variation in pitch. Early audio engineers heard similar perceptual blending when weak signals were dithered with noise to avoid quantization distortion in digital audio. The image processing sense adopted the name because the same principle applies: noise and deliberate imperfection create smoother perceived output than crude rounding.