Hello,
I open a noisy (6400 ISO) RAW image in Darktable 3.2 and enable the noise (profiled) module in “wavelet auto” mode. Setting the strength parameter to higher values has very noticeable effects.
Say that I set strength to 4 and play with the Y0 and U0V0 curves. Modifying the Y0 curve has a clearly visible effect, but I do not see any effect (at 200% zoom) of modifying the U0V0 curve except when I pull it down completely which seems to completely disable wavelet filtering of the U0V0 channel.
When, on the other hand, the U0V0 spline is just the slightest bit above the minimal value, resetting it to neutral by double clicking does not seem to have any visible effect.
Would someone who understands the wavelet denoise tool care to explain?
Thanks!
Hello
The curve internally has the same effect as in RGB mode: it affects the threshold that is used to denoise the wavelets. The wavelets coefficients are shrinked towards 0 with a soft threshold.
I don’t really know why it behaves this differently between RGB and Y0 and U0V0. I guess it is maybe due to the fact that U0 and V0 channels should have very low amplitude due to the way they are built from RGB data (which is not the case of Y0):
if all values are small, adjusting the threshold won’t impact much the result until the threshold reaches a value really close to 0 (having the curve all the way down)
Thanks a lot for the explanation, @rawfiner. Would it make sense to adjust the y-axis range of the spline for the U0V0 curve?
Yes, I guess so
Maybe it would be nice to have a non-linear range for reducing the curve