Correct me if I am wrong, but isn’t capture sharpening correcting real physical losses that happen between the subject and the sensor? Either way, the sharpness it applies seems different visual wise, than for example D&S. It almost feels like an increase in resolution as opposed to increased visual acuity that the other modules provide, if that makes sense.
Or use DoF while capturing 
I’d urge you to try it first. In my limited testing there are details.it brings out that other sharpening methods do not.
Generally, I do agree and the sharpening of my Z primes and gfx lenses is used with a light touch.
I tried it, it is great, I love it, but (at the moment) I think that I need to use contrast eq to undo it selectively, it is subtle but, on the whole image, is very noticable (in a good way). Maybe as I experiment with it more my attitude will change. I would describe the effect as upgrading to a more expensive lens line.
Glad you’ve tried it. I’m liking it a lot for my subject matter: peeling paint, old buildings, stuff with lots of texture. It gives it an extra little oopmh that I enjoy. Of course if you don’t find it suitable, then…
I’m excited to try capture sharpening (and eventually agx). I know this isn’t the point, but kind of funny how much the industry and consumers get invested in making the sharpest cameras, lenses and processing as possible only so they have to undo it again with a bunch other products. Makes me think of Japan Tobacco’s investments in cancer medicine.
To reemphasise, I’m just saying this in jest and I’m really happy about the new module and all the work other people put into it… 
My understanding from reading the code is that it is a small-radius, auto-tuning deconvolution-style algorithm, for which special care has been taking not to amplify noise (which is usually a weak point of naive implementations). Yes, it is definitely different from D&S which may span the entire image and operate at much lower frequencies. My understanding is that it is intended for small-scale effects, like AA filters, lens aberrations, etc. The neatest part is that you don’t have to specify how the unsharpness arose. It is a very clever algorithm.
All the cool YouTube kids are turning down the jpeg sharpening and shooting with a bloom/black mist filter.
Of course once you can have a thing, the trend goes back to thing -1.
Polaroid is dead, long live Polaroid. Lenses are super freaking sharp, long live the bloom filter. Film is dead, film is not dead. Etc. 
While this may be true, note that this (= global desharpening) is not what I suggested above. The point is to introduce a difference in sharpness (and other things): use capture sharpening and all the other tools at your disposal, just do it selectively.
However, because you cannot mask demosaic, the only practical way is to undo it later. Again, selectively, outside your subject.
There is a contrast threshold for the sharpening so its maybe not a mask but it is a way to impact where the sharpening is applied…you can display this as well
Yes, and here it is in action:
The image on the left shows how you can ensure the blurry background doesn’t get sharpened too much, and only the monkey’s fur is sharpened.
I’ve just watched your video episode 56 ‘explanation of the channel mixer’. I learnt more from this than in any other video such is my ignorance. This video should be a fundamental watch for anyone editing photos. So thank you Boris!
Yes, and you can combine that with drawn shapes and various parameters (using the detail levels slider in mask refinement to restrict the selection).
I think we were talking about CS in the demosaic module…I’m not at my PC but I don’t think it has any masking options??
There is no masking options for CS in demosaic module. And there won’t be any masking options as we are talking about demosaicing.
I agree that demosaicing (reconstruct a full color image from incomplete color samples) has nothing to do with masking options, since one wants color reconstruction to act on the complete set of sensor data reconstructing color for the whole frame.
I do understand that, as a system concept, capture sharpening (CS) is considered a global operation to counteract the inherent softness introduced during the capture process. However CS can have masking options in some form. There is nothing technical that would prohibit that. If we restrict CS discussion to RL-deconvolution we operate blindly anyway (point spread function is unknown and has to be estimated) and several parameters are added to the process so user can adapt. I view a mask, in this case, as just another parameter to adapt CS.
I must say that the capture sharpen option introduced in demosaic module works very well and it improves my pictures compared to not having the module. I am fine with the current.
But CS is integrated in the demosaic module in DT, and that module has no masking options at all. Adding them just for the CS operation may be non-trivial… (then again, CS is not in 5.2.1, so still only in dev versions)
