Error Diffusion Vs Ordered Dithering

Error-diffusion algorithms process the image pixel by pixel. When a pixel is rounded to the nearest available value, the leftover difference — the error — is distributed to nearby pixels that have not been processed yet. Floyd–Steinberg spreads that error to four neighbours, while Jarvis–Judice–Ninke and Stucki spread it across a wider area, which tends to produce smoother gradients with less visible structure.

The advantage is organic, photographic detail with no obvious repeating pattern. The trade-off is that the result depends on the exact pixels, so small changes between video frames can cause the texture to shift or shimmer.

Ordered dithering compares each pixel against a value from a fixed threshold matrix (often called a Bayer matrix). Because the matrix is the same everywhere, the resulting pattern is regular and repeatable. This gives a clean, screen-printed or retro look and, importantly, stays stable from frame to frame in video because it does not depend on neighbouring pixels.

The downside is that ordered dithering can look mechanical on photographic subjects, especially skin and subtle gradients. Raising contrast or adjusting the grid size usually helps it read as a deliberate graphic style rather than a limitation.

There is no universally correct option — it depends on the result you want. If you are converting a portrait or a richly detailed photo and want it to feel natural, start with error diffusion. If you want a controlled, graphic pattern, are working with flat illustration, or are exporting motion, start with ordered dithering.

Because switching modes in the workbench is instant, the fastest way to decide is to try both on your actual image and compare them in the split preview at the size you intend to publish.