NegPy version / build
NegPy v0.58.0
OS
Mac OS 26.6.2
Scanner
Digital camera (macro/copy setup)
Scanner model / details
D800 NEF
Light source
Big Scanlight, Trichrome camera scan Narrowband + LinearRaw, no IR plane.
Correction mode
On (narrowband trichrome correction enabled)
Input ICC profile (if any)
NA
Relevant process-tab settings
NA
What happened
With Optical Removal on and Overlay: Marked, thin high-contrast edges that are part of the image get marked magenta (structure-following fill) and repaired out. Shape doesn't matter: straight court lines and window mullions are caught, and so are curved leaf rims. It looks driven by edge contrast, not geometry. Checked at 1:1 with the Overlay off: no trichrome registration fringing on these edges, and Overlay: IR shows nothing, since there's no IR plane.
- Frame 170, tennis court: every white court line, plus a roofline. Threshold 0.32 and 0.96, Size 4 px. Same result at both.
- Frame 166, window mullions against a bright window. Threshold 0.96, Size 4 px.
- Frame 139, foliage: bright leaf rims marked magenta, leaf texture marked green. Threshold 1.00, Size 3 px, the most conservative the sliders go. Repair readout: painted 0.13%.
What looks off:
- Threshold has little effect on line detections. Same result from 0.32 to 1.00.
- The court lines are several times wider than the 4 px Size cap but still get caught. Anything line-shaped seems to bypass Size and go to the wide fill.
- Frame 139 is busy detail, where the docs say the threshold rises. It still catches edges at max threshold.
Expected: real image lines and edges survive at conservative settings. An exclusion mask, or tying max line width to Size, would fix it.
Suggestion: split Optical Removal detection into two checkboxes, "Spots" and "Lines/Hairs", both on by default. On frames with strong image edges you could run spot removal alone and handle the occasional hair with the Scratch or Transport Line tools. Better line discrimination is the real fix, but this would make auto dust removal usable on those frames in the meantime.
Screenshots of before/after attached, along with links to triplets.
Source file
Frame 170 an d139: https://SchweiNas.synology.me/SchweiDrive/d/s/19paByP7mYJ7UOlhabpr1doNVMrdd9lT/B75js-vaUijNFfPR1Lj28HJzietU7rpQ-0LYg5BGlfg0
Screenshots
Frame 170
Frame 139 (Overlay Off / Overlay Marked / Optical Removal on)

NegPy version / build
NegPy v0.58.0
OS
Mac OS 26.6.2
Scanner
Digital camera (macro/copy setup)
Scanner model / details
D800 NEF
Light source
Big Scanlight, Trichrome camera scan Narrowband + LinearRaw, no IR plane.
Correction mode
On (narrowband trichrome correction enabled)
Input ICC profile (if any)
NA
Relevant process-tab settings
NA
What happened
With Optical Removal on and Overlay: Marked, thin high-contrast edges that are part of the image get marked magenta (structure-following fill) and repaired out. Shape doesn't matter: straight court lines and window mullions are caught, and so are curved leaf rims. It looks driven by edge contrast, not geometry. Checked at 1:1 with the Overlay off: no trichrome registration fringing on these edges, and Overlay: IR shows nothing, since there's no IR plane.
What looks off:
Expected: real image lines and edges survive at conservative settings. An exclusion mask, or tying max line width to Size, would fix it.
Suggestion: split Optical Removal detection into two checkboxes, "Spots" and "Lines/Hairs", both on by default. On frames with strong image edges you could run spot removal alone and handle the occasional hair with the Scratch or Transport Line tools. Better line discrimination is the real fix, but this would make auto dust removal usable on those frames in the meantime.
Screenshots of before/after attached, along with links to triplets.
Source file
Frame 170 an d139: https://SchweiNas.synology.me/SchweiDrive/d/s/19paByP7mYJ7UOlhabpr1doNVMrdd9lT/B75js-vaUijNFfPR1Lj28HJzietU7rpQ-0LYg5BGlfg0
Screenshots
Frame 170
Frame 139 (Overlay Off / Overlay Marked / Optical Removal on)