I notice that within the processing order “denoise (profiled)” occurs very early in the pipeline and before most of the processing takes place. This appears to me as a little odd.
Is there a reason for this? I thought that this module was a correction aimed at processing noise.
1 Like
So isn’t that logical. Why would you want to process noise. In fact many modules might be negatively impacted by noise… You have other thoughts??
My understanding is that raw-denoise’ takes care of the camera’s inherent noise and then the profile-denoise is an attempt to remove noise created by modules during the course of the processing. Which would then come much later in the pipeline.
Not sure where you got that idea from. The manual is pretty clear that it’s for camera noise: darktable 4.6 user manual - denoise (profiled)
An easy to use and highly efficient denoise module, adapted to the individual noise profiles of a wide range of camera sensors.
Also, if you have properly denoised the image early in the pixelpipe, later modules shouldn’t introduce much, if any, noise. But if you do have noise then, the best tools are probably contrast eq and d&s.
Obviously, I was unaware of the source of noise in my images. Most noise that I viewed appeared to emulate from processing and appeared to be ‘fixed’ by careful application of the profile-denoise.
I guess I need to rethink my processing in some areas.
By the way … I do not only read the manual but also print a new copy at each major revision.
Hi David,
I find Raw denoise softens the image a lot and I rarely use it. The denoise profile tends to work great for most images. Noise can mess with filmic’s calculations if you use the auto pickers so removing it early makes sense.
2 Likes
I think that the module uses noise profiles for camera iso and any lens combination that you might have profiled…
Explained here
So in essence it is using a noise profile of the sensor at a given iso to help correct noise. Noise is subsequently magnified by each processing step so introducing it late would sort of defeat the purpose and effectiveness in the end… at least that is how I understand it…
All the methods are very well explained over 3 videos by a YT creator called rawfiner. He is the author of the module but he demonstrates all the ways that you can denoise in DT… they are excellent videos including one more that he made on the CA module that he authored as well…
This video covers tweaking raw denoise and all of the other methods in DT… its from 2.6 but still very applicable as are the follow up ones for 3.0 and 3.01 where the DN profiled module was further modified… also the Chromatic Aberrations video is a good one if you have not seen them…
1 Like
Erm, how would processing modules introduce noise (= random variations in the signal)?
Certain modules can make it more visible, true, but that’s not the same thing.
(Of course, a module can be designed to add noise, but that’s intentional)
As modules not designed to introduce noise, don’t add any, but can amplify existing noise, it makes sense to denoise as early in the pipeline as possible.
Also, lenses shouldn’t introduce noise…
A bit more techical: most modules treat noise as having a gaussian distribution. As the sensor signal can’t go negative (in essence, you’re counting captured photons), that can’t be correct. Such noise should be treated as a mixture of a Poisson distribution and a Gaussian (the Gaussian part comes from the electronics.
This is what “profiled denoise” does: the profile describes the relative contributions of the Poisson and Gaussian distributions). As the Gaussian part can vary between sensor models and electronics, you want camera-specific profiles.
But (like any noise reduction) it slightly blurs the image (you’re averaging pixel values over a small area, after all).
3 Likes