Starting with capture sharpening, I agree that the main intention is to restore information lost in capture. So radius would be chosen to match diffraction + OLP filtering and deconvolution would be used. But, that would not be the optimal setting for portraits as it enhances both real and fake detail.
Hi Erik,
It's important to note, and it's not just semantics, that
proper Capture sharpening doesn't 'enhance' real detail, in the sense of magnifying the amount, it merely attempts to restore it to what it was before it entered the lens. What better starting point for postprocessing could one have? We can deal with skin pores and blemishes by using the proper tools for that, or modify the visibility of original detail with a lens filter and lighting.
If the Capture process also generates fake detail, photon shot noise and or grain, aliasing artifacts, dust bunnies, etc., then we need to deal with those with the proper tools, but not with (what's more akin to) Creative 'sharpening' that disrupts the restoration of the original detail from the subject.
Good deconvolution tools attempt to restore resolution, but not amplify noise, or at least restore more resolution than amplify noise for a better S/N ratio. We can use noise reduction in that process, as long as the noise reduction doesn't blur the image too much, because that would require increasing the deconvolution radius. Cascading blurs amount to a larger blur radius, because one can create a large radius blur by cascading multiple smaller blur radii.
This is where it gets interesting, because some noise reduction programs use a combination of blurring, and edge detail preserving techniques, which breaks the original signal's true composition of spatial frequencies. Such noise reduction would make it much harder to restore the original detail, because it was changed into something artificial that doesn't comply with the laws of physics anymore.
That's why a good deconvolution method tries to separate the noise from the signal, and only restore the latter!
If the sharpening adds artefacts of it's own, it is reasonable that it would be done at the largest image size in the process.
Yes, if e.g. stairstepping, or even worse halos, are created by the Capture sharpening process then we are making life only more difficult for when we want to generate large output. However, the more accurate we choose our Capture sharpening settings, the lower the risk of creating 'detail' which was not in the original signal. Again, with Capture sharpening we strive for the restoration of original subject detail, nothing more, nothing less.
The way I handle it, right now, is that I use a small radius (0.7) and relatively large amount with deconvolution in LR. If I stop down to f/16 or more I increase radius to 1.3. This works well for my standard print size which is A2.
Yes, as you saw from my charts earlier, a radius of 0.7 is about as good as it gets, and that's true with most lens/sensor combinations (assuming a Raw conversion without prior sharpening), and when diffraction kicks in, we (or rather the Capture sharpening tools) need to work a lot harder to restore the original signal. But a lot is possible as was demonstrated in the
Deconvolution sharpening thread here on LuLa, with examples
here, as long as we use the correct settings.
The moment we deviate from proper
restoration to 'enhancement', we tend to create trouble for ourselves later in postprocessing.
Cheers,
Bart