Dithering for Laser Engraving: A Practical Guide

What dithering does

Most lasers do not reproduce a photo in the same way a display shows it. A common engraving workflow treats each output position as mark or no mark. Dithering converts the photo's continuous range of tones into a binary pattern whose local density creates the visual impression of gray.

Viewed from normal distance, a dense field of marks appears darker and a sparse field appears lighter. Viewed at 1:1, the file should contain only the exact black and white pixels that will drive the pattern.

Sources: 1, 2

Error diffusion, ordered patterns, and threshold

Error-diffusion methods decide whether a pixel becomes black or white, then pass the difference between the original and chosen value into neighboring pixels. This produces irregular, photo-like dot fields. Floyd-Steinberg uses a compact neighborhood. Jarvis, Stucki, and related methods spread error more broadly, which changes smoothness and edge behavior.

Ordered dithering compares pixels against a repeating threshold matrix. The regular structure can work well for graphic shapes and broad fills, but the pattern may be visible in photographs. Halftone and newsprint styles make the pattern more intentional by changing spot size or arrangement.

Threshold is not a general photo dither. It sends every pixel to black or white based on one cutoff. It is useful for artwork that is already two-color, or for a prepared one-bit bitmap that needs to remain unchanged.

Sources: 1

How to compare common algorithms

Floyd-Steinberg is a useful baseline because it is compact and widely understood. Jarvis often creates smoother tonal transitions at the cost of a broader dot field. Stucki has a similar photographic use and can retain a slightly different balance in midtones. Atkinson passes less total error onward, which can produce a lighter, crisper result while losing detail near the ends of the tonal range. Bayer ordered patterns are regular and predictable, which can be helpful for graphic work or when a visible texture is acceptable.

Those descriptions are starting points, not material laws. The final appearance depends on the prepared tones, output resolution, laser spot, scan spacing, heat spread, and surface response. Compare methods on the same crop without changing other controls.

• Portraits: inspect skin transitions, eyes, hair, and the boundary between face and background.
• Pets: inspect fine fur, bright whiskers, dark eyes, and patterned coats.
• Architecture: inspect thin edges, repeating textures, skies, and shadowed windows.
• Logos and line art: consider threshold or ordered patterns before a photographic diffusion dither.

Sources: 1

Judge the dither at final size

Dithering before final sizing is risky because resizing a binary pattern is not a neutral operation. Downsampling can erase dots or merge them. Upsampling can duplicate pixels or introduce interpolation. Either changes the relationship between the prepared pattern and the laser's scan grid.

Set crop, physical dimensions, and intended DPI first. Then dither the resulting pixel grid. Use a fit view to judge composition and a true 1:1 view to judge the marks themselves.

Sources: 1, 3

Fix the tonal map before blaming the algorithm

A dither can only distribute the tone information it receives. If the face and background occupy nearly the same tonal range, switching from Stucki to Jarvis will not create strong separation. Adjust exposure, contrast, gamma, local detail, and background treatment before making fine algorithm comparisons.

Watch the histogram conceptually even when the app does not show one. A photo packed into a narrow middle range tends to produce a busy field of similar density. A photo with all shadows clipped tends to produce large solid regions. Preserve meaningful separation without forcing every image to pure black and white before the dither stage.

Sources: 4

Do not dither twice

Double dithering happens when a prepared one-bit image is imported into software that treats it as a new grayscale photo and applies another image mode. Resampling can create the same problem even if the second step is not labeled dither.

The symptoms include moiré-like patterns, missing fine dots, large dark patches, unexpected softness, or a preview that no longer matches the exported bitmap. Use the destination's pass-through or equivalent preserve-pixels mode, keep the intended physical size, and verify the preview.

Sources: 1

Run a useful dither test strip

A good strip uses one representative crop and changes only the dither method. Include texture, a highlight, a midtone, a deep shadow, and an important edge. Label each variation outside the image area so the mark pattern is not confused with the label.

First compare which files preserve the subject. Then run the strongest two or three on scrap using the same machine settings. This separates image preference from physical material response and reduces the temptation to change five variables at once.

Sources: 2, 3