Thank you, very instructive and useful.
When you mention the method to make mask with ACDNR, you also state that ACDNR allows smoothing the mask (sldie 48).
Should the mask not be smoothed (that is denoised) systematically ? I see often that a simple intensity stretch of the L image is used as mask for the first denoising. Does the noise of the mask not get applied to the image via the masking effect ? My understanding is that the mask provide a mixing factor between the original and the denoised image, and if this factor is itself noisy, this is somewhat reflected in the result.
I think I had slightly bettter results with a smoothed mask when working on quite noisy images. But I do not know if this effect would be important or event significant on less noisy raws.
What do you recommend?
-- bitli