Start by turning on the color equalizer’s saturation-weight visualization: it shows which pixels may be affected according to saturation, before you decide whether to raise the threshold. Then make one change at a time. A higher threshold restricts the operation on less-saturated pixels; sharper hue-curve interpolation narrows the hue transition but can increase artifacts; larger guided-filter radii can smooth changes yet may cause edge bleeding or halos.
Diagnose in order
- Select the hue, saturation, or brightness tab you intend to adjust, and make a restrained curve change. Use the picker to locate a sampled color on the curve if that helps identify the intended hue. Treat node colors as a visual clue, not a control label: check the actual tab and adjustment type.
- Enable the saturation-weight visualization beside saturation threshold. Look for low-saturation areas that would be included in the edit, such as neutral fabric or colored speckles. This view depends on saturation and the contrast setting, not on the adjustment curves. It is a scope preview, not proof of the final effect.
- If unwanted low-saturation pixels are included, raise the threshold slightly and compare the preview. If the transition from affected to unaffected pixels seems too abrupt, assess the contrast control as well: darktable documents it as changing the steepness of the saturation-weight curve. Do not alter both together in the first comparison.
- If noise or artifacts follow the hue adjustment itself, inspect hue curve interpolation. Values below 1 create sharper transitions between nodes and affect neighboring colors less, but raise the risk of noise or artifacts. Values above 1 make transitions more gradual. Try a less sharp setting before increasing filter radii when the target boundary is the suspected cause.
- If introduced noise remains, increase hue analysis radius in small increments. This includes more neighboring pixels in hue analysis and may smooth noise, but very large values can create halos at edges. Evaluate effect radius separately: a larger value smooths the applied adjustments and can retain existing detail, but may cause bleeding at edges. Use the second visualization beside effect radius to inspect changed output for the selected tab; it reflects both the curve and guided filter.
The manual describes the guided filter as a compromise between color-targeting precision and reducing artifacts. Its color equalizer reference explains these controls and visualizations. Do not assume a wider analysis radius will fix every halo: it can introduce one, while effect radius has its own edge-bleeding trade-off.
Hypothetical acceptance worksheet
The observations below are invented diagnostic examples, not measured darktable results or guaranteed outcomes. Compare the same crop and adjustment at the same zoom; this is an editorial comparison method, not a feature or measurement rule defined by darktable. Record the starting settings and change only the named control between comparisons.
| Hypothetical detail | Weight-preview observation | Curve and radius check | Next single adjustment to compare |
|---|---|---|---|
| Fine patterned fabric | Blue and red both appear across low-saturation threads and colored pattern | Target hue curve is sharp; analysis radius is small | Raise saturation threshold slightly; then inspect whether the intended threads remain included |
| Colored noise in a muted area | Speckles appear among pixels that receive some weight | Hue curve is below 1; effect radius is unchanged | First compare a less sharp hue interpolation; if noise persists, increase analysis radius incrementally |
| High-contrast leaf edge | Preview shows the leaf is weighted; the edge looks cleaner, but a colored fringe appears | Analysis radius has been enlarged | Reduce analysis radius and compare; do not keep enlarging it to treat the halo |
| Bleeding beyond the leaf | Fringe appears in the changed-output visualization | Effect radius was enlarged while analysis radius stayed fixed | Reduce effect radius and compare the selected tab’s visualization |
Accept a revision only when the intended color change remains useful and the crop no longer shows an unacceptable new artifact. Note which control changed and what happened; a cleaner preview alone does not establish that the rendered image is acceptable.
Limits and next steps
The saturation visualization indicates possible impact based on saturation; it does not account for the hue adjustment curve. The changed-output visualization, by contrast, reflects the current tab’s curve and guided filter. Use them for different questions: “which pixels may be weighted?” and “where does this tab change the output?” The official options documentation also warns that noise and artifacts may not be fully eliminated.
If raising the threshold removes too much of the intended color, return to the previous setting and compare a gentler hue curve or smaller curve change. If analysis-radius changes create halos, reduce that radius; if bleeding follows effect radius, reduce that control instead. If no restrained setting gives an acceptable result, reduce or remove the color equalizer correction for that area rather than claiming the filter restores detail. The manual does not promise a universal best radius or threshold; the useful choice depends on the image and the trade-off visible in the crop.
Common questions
Does the saturation preview update when I reshape the hue curve?
No. It represents saturation weighting and is affected by contrast, not by the hue, saturation, or brightness adjustment curves.
Should I turn off the guided filter to eliminate artifacts?
Not as a default fix. The manual says it is enabled by default to reduce noise and artifacts. Turning it off removes that filtering compromise rather than guaranteeing a cleaner result; compare cautiously if you choose to inspect it.
Are hue analysis radius and effect radius interchangeable?
No. Analysis radius changes how neighboring pixels contribute to hue analysis; effect radius changes the spatial smoothing of applied adjustments. Diagnose which change coincides with the visible problem before adjusting either.
Sources and next step
This guide interprets public documentation and uses explicitly hypothetical acceptance examples, not application tests. Controls and availability can change; verify the cited manual against your installed version.
For work created in Imagild, carry the input requirements and acceptance checklist into the photo editor. The external software controls discussed above belong to their respective products.