To strengthen mid-scale surface structure in darktable 5.6, start with the diffuse or sharpen module’s local contrast preset and adjust its scale range to the texture you want, rather than using pixel-level sharpening or increasing overall contrast. Compare the same crop before and after, especially in flat shadows and at sharp edges. If your aim is general display-referred local contrast rather than scene-referred scale selection, compare the separate local contrast module instead.
Choose a scale and make a controlled comparison
Use diffuse or sharpen when the target is a visible band of texture, such as folds in fabric, grain in wood, or broad stone texture. It operates on linear, RGB, scene-referred data. It is not simply a global clarity control: its local contrast preset works through diffusion across selected detail scales. The diffuse or sharpen manual describes central radius as the main scale and radius span as the range around it.
- Identify one mid-scale feature to strengthen. Do not make “sharper” the target: decide what surface structure should become easier to read.
- In Darkroom, compare the photo with the local contrast preset enabled against the unchanged version. Keep the crop and zoom consistent so you are comparing the same structures.
- Set the central radius with the target detail in mind. The manual says non-zero values select the detail size affected more heavily; the radii refer to pixels in the full-resolution image. Reassess after changing image resolution, since settings may produce different results at different resolutions. If you only change the crop or viewing area, check that the target texture remains within the affected range; do not treat cropping as a resolution change.
- Keep radius span focused. For local-contrast enhancement when you do not want to affect sharpness or noise, the manual recommends setting radius span to three quarters of the maximum value available for central radius, not three quarters of the central radius value you selected. Treat this as guidance, not a guarantee that a particular image will be noise-free.
- Inspect a flat, dark area and a high-contrast boundary. If noise becomes conspicuous, the intended surface structure is overwhelmed, or edge fringes appear, reduce the effect or revisit the scale range. The manual identifies edge sensitivity as a control for reducing edge artifacts; change it only when the edge evidence points to that problem.
Avoid substituting the sharpness control for this scale choice. The documentation warns that using sharpness alone for sharpening can produce halos or fringes. Nor is this a deblurring workflow: deblurring uses different scale guidance and can magnify existing noise.
Record what the comparison actually shows
This author-proposed worksheet is a hypothetical acceptance example, not a software test. The scale descriptions follow the manual; use the observation prompts on your own image.
| Hypothetical target and observation | Preset and scale record | Decision |
|---|---|---|
| Wood grain reads more clearly at normal view; dark tabletop looks unchanged | Local contrast; record central radius and span shown in the module | Keep only if the grain, not a crunchy edge, is the visible improvement |
| A shadow crop shows more colored speckling than the baseline | Same settings and matching crop; do not infer a universal noise threshold | Reduce the effect or reconsider the target scale |
| A bright edge develops a visible fringe | Record the scale settings and edge sensitivity used | Back off; test edge sensitivity only if the fringe remains the relevant defect |
| After changing crop or working resolution, the texture looks different | Recheck full-resolution pixel radii against the new image | Retune by observation rather than copying the old numbers blindly |
Write down the preset name and the two radius settings alongside each comparison. That record makes the next decision about scale, not merely whether the image looks “stronger.”
When to switch routes or stop
Choose local contrast when you want a different kind of control: it works on the L channel in Lab in a non-linear, display-referred stage, using a local Laplacian filter or bilateral grid. Its controls include detail and, in local Laplacian mode, shadows, highlights, and mid-tone range. Those controls are not equivalents of diffuse or sharpen’s central radius and radius span; the modules operate at different stages and offer different ways to shape contrast.
If the target texture only becomes visible by making shadows noisy or edges fringed, do not keep widening the scale band. Return to a subtler setting, consider local contrast if its display-referred controls suit the job, or accept less enhancement. The manual describes diffuse or sharpen as resource-intensive and says results are guaranteed only at 100% zoom and high-quality or full-size export; results at lower zoom or export dimensions may differ. Check the intended viewing or output size.
Common questions
Does a non-zero central radius mean pixel sharpening?
No. The manual associates zero with stronger action on fine details, while non-zero values select a detail size for local contrast. Both values are interpreted in relation to full-resolution image pixels.
Does the suggested span prevent noise from increasing?
No. The manual gives it as guidance for local-contrast scale selection, not as a guarantee against noise. Check a flat shadow in your image.
Should I use local contrast for every mid-scale texture?
Not automatically. Use diffuse or sharpen when its scene-referred diffusion scale selection matches the task. Compare local contrast when you want its Lab, display-referred approach and its detail or tonal controls instead.
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.