I’m relatively new to Darktable and raw processing in general. I was testing it out on a challenging file today.
After using the Haze Removal module, I noticed excess colour noise, particularly in the centre of the image. Maybe I’m pushing this too hard or expecting too much from this module?
I intended on tackling this by using the Denoise Profiled module with wavelets and strength set to 1.
When I applied this, it seemed to mess around with the highlight reconstruction and color reconstruction on the sun. I tried applying the Denoise module with a parametric mask to exclude the brightest pixels, but for some reason, it seems to be ignoring this and applying the denoise to the whole image. I have also tried using a drawn mask to exclude the area with the sun to no avail.
Any help much appreciated!
Here is a screenshot with the denoise module disabled:
strange effect indeed. Looks as if masking doesn’t work for the raw-clipped regions of the image.
If you crank up the strength of denoising and watch a sharp border of the mask (set the sliders so there is no gradient) you can see that masking works for the unclipped parts of the image.
darktable moduls work in different color spaces. Conversion could create color artefacts in some cases. Maybe that is the cause? You could try to switch “gamut clipping” on in the input color module like descripted in the manual.
My pure guess is that turning on the module forces a color space conversion on the module input. The modul ouput is blended with the pixels of the mask. Outside the mask there is no blending of the pixels, but still the color space conversion is happening.
i can confirm this “bug”.
it only appears when you use negative values in exposure module or other modules and color or lightness blending in denoise module
try with blend mode HSV Color or HSV Lightness in denoise module.
or exclude the highlights in exposure module, haze removal and color reconstruction with parametric mask.
.