Modules which operate in Lab may create visible artifacts even when configured as no-op (all sliders at zero / straight curves)
I've found some problem samples which cause certain modules ("contrast brightness saturation", "tone curve" when not in RGB or XYZ mode) cause visible artifacts (violet dot clusters in this case) when enabled, even if they are configured to do nothing (which is the default).
Steps to reproduce:
1: Open the raw sample in darkroom mode, optionally use the supplied minimal xmp profile (it has only base curve and exposure and nothing else)
2: Enable "contrast brightness saturation" or "tone curve" with default(no-op) settings
Then look at the changes in blue areas
I also tested the previous release (2.4.0), there's no difference.
Also, the artifacts are not visible when gamut clipping in "input profile" is enabled, but this is a crude workaround, I'd rather not have to simply throw away potentially useful information early in processing pipe (if something is out of gamut, it can be brought in by reducing saturation or something else, am I right?).
I am also attaching the in-camera JPEG for reference. (reduced by jpegtran, so no metadata, but it can be read from the raw file if needed)
Right now I have 2 more similar raw files which exhibit the same problem, feel free to ping me if I should upload them too, or do something else to help debugging this.
#4 Updated by Sarge Borsch about 1 year ago
- File screenshot_nothing.png added
- File screenshot_gamut_clipping.png added
- File screenshot_cbs.png added
Roman Lebedev wrote:
Please attach both screenshots.
Looking at this screenshot, isn't this again gamut clipping issue?
Try setting "gamut clipping" (!! NOT the color profile itself) in "input color profile" to "rec2020"
Done. Also, tried "gamut clipping" with "rec2020", it removes the difference but it itself visibly changes the image in a different way. I'm attaching all 3 screenshots.
#6 Updated by Roman Lebedev about 1 year ago
Sarge Borsch wrote:
Roman Lebedev wrote:
Tried locally, it is standard "deep saturated blues" problem, which is described in usermanual.
Setting gamut clipping to rec2020 helps.
Shouldn't it be the default setting then?
Not really, it isn't essential/required to get some base image to be displayed.
Also, adds some memory/computational overhead.
You can, as usual, create auto appliable preset with whatever params you want.