dither

Dither is a technique used to minimize perceptual artifacts that arise when a signal is quantized or otherwise reduced in precision. By introducing controlled noise or by distributing quantization error over neighboring samples or pixels, dithering makes the result appear smoother and more natural to the human eye or ear.

In digital imaging and color quantization, dithering helps reproduce a broader range of colors than a limited palette would otherwise allow. This is achieved by creating small, patterned differences that, when viewed from a distance, blend to produce the impression of intermediate tones. Common methods include ordered dithering, which uses a fixed threshold matrix, and error diffusion, which distributes the quantization error to surrounding pixels. The Floyd–Steinberg algorithm is a widely used form of error diffusion, while Bayer matrices illustrate the classic ordered approach. Variants such as blue-noise dithering aim to minimize certain visual artifacts.

In audio, dithering involves adding very low-level noise to a signal before quantization to decorrelate the quantization error from the signal itself. This reduces audible distortion when lowering bit depth, especially for quiet passages, and is typically applied at the final stage of digital-to-analog conversion or during mastering.

Dithering is also important in printing and displays, where halftoning and dithering patterns simulate intermediate colors and gradients. This helps mitigate banding and improves perceived color and tonal transitions in images, video, and printed output.

Historically, dithering concepts emerged in mid-20th-century computer graphics and printing, with notable algorithms such as Bayer’s ordered dithering (early 1970s) and Floyd–Steinberg error diffusion (1976). Today dithering remains a standard tool in image processing, audio encoding, and display technology.