I don’t have a handle on the settings of denoise (profiled). Most of the time, I just enable it and move on.
This image is ISO 1000 with a significant crop. Noise really shows up. I understand part of the reason is that I under-exposed by so much (-2.3 ev) to avoid clipping highlights.
In denoise(profiled), I made extreme changes in the detail scale by brute force and came up with this. Chroma noise removal didn’t seem to cause much blurring, so I took it to the max on all scales. On luma, it seemed to cause a lot of blurring by denoising the coarse details, but not much the fine details.
I was “successfully” able to keep detail in the bird’s feathers. There’s still a lot of noise in the leaves, though. Yeah, I can fiddle with the strength and preserve shadows sliders, but I’m already beyond my understanding.
Questions:
How can you tell luma noise from chroma noise?
What are the fine and coarse scales?
Are the fine and coarse scales resolution-independent (dependent on source resolution? Crop vs. not cropped?)
Have you watched rawfiners videos on denoising and the module…He is the module author and has set of videos that are older now but still accurate explanations of noise and the DT tools… Even his video from version 2.6 has useful information on how to visualize noise…channel by channel…I would check them out
Often just moving the preserve shadows slider is a good tweak on the default of profiled denoise. I also often will use only the chroma preset and then use astro denoise or the contrast eq for Luma noise. I just depends which pairing works best in a given instance…
Lost in that you don’t know what I am referring to??? If so I was referring to the chroma preset of denoise profiled and then to try astro denoise and or the contrast eq for Luma denoising… is what I meant… surface blur which is the old bilateral denoising can also be effective and of course masking can really help as well to preserve detail in the bird…
Here is your image…I just bumped up the vibrance and chroma a bit to exaggerate it…
In DT you can see the fine and course wavelet scales in the retouch module. You could also blend your denoise module in difference mode and play with the various scales and see what gets impacted…
A nice video to demonstrate it as well is here…its an old video about the contrast eq but basically I think this is the same wavelet math or very similar used in the denoise wavelets and diffuse or sharpen…in denoise there is no setting for radius but you can use the scroll wheel…
The explanation is around 5 min into the video… to this day this is also one of the best contrast eq video’s…
The last question I am not sure of but I don’t think the crop impacts the denoising or likely the wavelets as it happens before cropping in the pipeline… I suppose the crop might impact your visual assessment of things but if you look at 100% or with the HQR preview I think that would normalize things…
Here is one quick version I did with profiled DN and surface blur. I also got a decent version messing around with nlm version of profiled DN as well…
I was trying to understand how the graph of wavelet mode in denoise (profiled) correlates with the image. How do you decide whether to raise / lower the coarse / fine noise for the luma / chroma channels? I don’t understand the different kinds of noise or how to adjust denoise (profiled) to affect the “target” noise.
Rawfiners video for version 2.6 shows a basic analysis illustrating how to visualize the noise…fine/course for luma and chroma…The curve just represents a spectrum from course to fine so moving a selected part of the curve will impact or focus the adjustment on that scale of noise. You can blend the module in difference mode to see exactly what you are impacting while you experiment
For noisy images I find LLMSE (with colour smoothing) works better than RCD debayering. Used wavelets denoise masked to affect the lower luminances more, stacked with a non-local means denoising instance masked to the background - this removed most of the noise except for the very fine ‘texture’ like noise. Finally I used a diffuse and sharpen instance with the ‘denoise: fine’ preset to remove that fine noise.
I used to try and remove all of the noise in the profile denoise module, but I found it would also remove all of the detail in doing so, hence the multi-step approach.
