When working from a jpeg source (no RAW available) and trying to reduce noise, the two things I am using is Surface Blur, and Denoise (profiled). Most often the latter.
With Denoise (profiled) there are two choices. I find “non-local means” can get cartoonish, and wavelets sometimes misses some noise. I think the results from Surface Blur are similar to wavelets, maybe a bit cleaner, but also maybe it loses some detail as you can’t control the strength. What do you think works best and how do you adjust it? I am a bit vague on what radius does, for example.
I haven’t pursued using Diffuse and Sharpen. Is that worthwhile? Are there other options for general denoising a jpeg?
Go to Rawfiners youtube site…he has about 3 video’s…one is actually dating back to version 2.6 of DT but it is a deep dive on basically most of the ways to denoise in DT and how to tweak them…surface blur is an updated module version that was previously called the bilateral denoise module…
The defaults for NLM can be harsh but they can also be dialed back and you can sometimes get a very good result…
“Denoise (profiled” uses certain assumptions about the distribution of the noise. A jpeg probably already has some denoiding done to it. And that means that those assumptions are not valid anymore.
You can try to adjust the curves in the wavelet denoise to tune the denoising?
As for controlling the strenght: you could try using a blend mode (“normal” with opacity <100%, or “chromaticity”/lightness). That method works for any module, btw.
The “contrast equaliser” has “denoise” preset, and you can adjust the curves to adapt the profile to the noise you have left in the image. It also has a “strenght” control: the “mix” slider.
And of course, with denoising you are always trying to find a balance between removing noise and keeping details.
If I was trying to denoise a JPG I would consider astrophoto, contrast equalizer and some of the presets in the diffuse or sharpen module. I also take snapshots of each step to do a comparison as the difference can be subtle.