I am re-considering my workflow. Reading up on sharpening, I stumble over this passage by Bart van der Wolf in an old thread:
http://www.luminous-landscape.com/forum/index.php?topic=54798.msg447163#msg447163
> What few people seem to understand is that the upsampling itself can be mathematically characterized by a Point Spread Function (PSF). That fact can be exploited by using deconvolution sharpening after upsampling. It can even be taken as far as skipping the capture sharpening, thus avoiding any risk of artifacts and enlarging those, and combining the capture sharpening with the output sharpening with a (somewhat timeconsuming) restoration of detail, called deconvolution at the native print resolution.
Hi Hening,
Sharpening has several aspects to it that are related to the stage of the wokflow one is at, and it can become quite involved (Jeff Schewe even co-authored an entire book on the subject), but it is the final result that counts. If we can avoid creating artifacts early on in the workflow, then there is no problem with applying Capture sharpening then. When we create output, we will need to address the relevant sharpening for that stage at that time. The matter that I raised is, that it may be possible to combine the two sharpening steps into one, which would have the benefit of not enlarging early artifact creation with the increase of output size.
I still like the concept of Capture / Creative / Output sharpening, because these three stages allow to focus on the relevant issues for that specific stage of the workflow. That also has to do with the limitations of the tools we use. For instance, if we do not have proper output sharpening tools, we need to do more earlier in the workflow so we can do less later. Again I would like to mention that Creative sharpening is more about local contrast enhancement than real sharpening, but it's easy to refer to them as three 'sharpening' steps, also because contrast affect the perceived sharpness (even if it is only due to contrast and not intrinsic sharpness).
This rises some questions to me.
a- It seems that the advice presupposes that the one sharpening is = the print sharpening, and that one prints oneself.
Would it also be advisable to delay the *screen* sharpening to the last step before sending the files to print, leaving the print sharpening to the print service?
I could in principle try to do that, sharpening in Qimage, then send the files to the service, asking him not to sharpen. I have tried that, but his result was better than mine. That was after prior capture sharpening (deconvolution in Raw Developer, now Iridient).
The issue is that many of the tools we use offer little guidance as to their optimal settings. That leaves us with running tests and visual inspection of the final results, and going back to the drawing board for another attempt. That's why I advocate a more analytical approach, to take the guess-work out of it, for a large part anyway. And since Capture sharpening is in general well suited for an analytical solution (unfortunately not offered in most Raw conversion and Capture sharpening tools), it's a good candidate. One element of Output sharpening, the "Resampling blur", also allows an analytical solution.
Mathematically those two could be combined, although things may get complicated if the Creative 'sharpening' in between has significant effects on the underlying Capture blur. Therefore, the simplest thing is to separate the three stages, but it also triples the effort. It does offer the most control, especially if we use good tools.
Qimage has an excellent Deep Focus Sharpening (DFS) capability for tackling the output sharpening, eventhough it is not Decovolution based. It therefore builds on the quality of prior sharpening steps. Deconvolution
Capture sharpening is therefore almost a requirement, but without the creation of artifacts.
b- The 'complex' workflow means it implies both exposure stacking and focus stacking. In this situation, it may seem that delaying the deconvolution will be in conflict with another advice, also from your pen (if memory serves me), namely to deconvolve as early in the pipeline as possible. This seems to make sense. Otherwise, unsharp stuff is fed to the stacking software.
The more complex a workflow gets, the more we can benefit from control over each step. Focus Stacking is special, because it requires very accurate management of the individual slice's DOF. When we have focus intervals with OOF zones between them, then Capture sharpening will not do much more than point out the flaws in our technique. Sharpening the stacked result, may give better results (if the stacking operation does a good job on the less sharp slices). So that's a bit of a Catch 22, it may also depend on the subject matter what will work best. Exposure stacking/fusion should not affect our sharpening approach much, since it's more about improving the signal to noise ratio which allows us more Creative 'sharpening' wiggle room (AKA tonemapping and local contrast adjustments).
So I see a choice of 3 workflows:
1- One sharpening in Qimage, trying to nail the print sharpening at once.
2- One *screen* sharpening at final print resolution after upsizing and re-sampling, still leaving the *print* sharpening to the service
3- Capture sharpening in the beginning, using upsizing-sharpening-downsizing. Then sending the files to print, enlarged but not resampled, leaving this and the final=print sharpening to the service.
By 'sharpening', I always mean deconvolution. Typical image content is high contrast and lots of fine detail (landscape). Print size is 70x100 cm.
That is all reasonable, depending on how well the different stages are executed. It seems a bit strange to me that the print service can do a better job of output sharpening, unless one does not have adequate feedback from them what the printing process requires. One needs to know the output specifications (PPI) and medium requirements (e.g. paper diffusion combined with the output process, e.g. inkjet or laser C-print). When the service treats images individually and they know what they are doing, it could be left to their expertise (which should not be taken for granted in general because they rarely have the time to tweak).
With a multi-stage sharpening workflow, nothing beats proper deconvolution Capture sharpening. However, to do it proper requires good tools. Most tools require tuning, more tuning than necessary if only the tools were implemented better. The current brain-dead approach of leaving it to visual judgement (at which most people are not very good) while the process could be guided by image analysis in the background still amazes me (from a technological point of view, I understand it from an investment/profit point of view).
Creative 'sharpening', e.g. with Topaz Clarity for local contrast and Topaz Detail for selective structural detail enhancement, is fully separate from the other less creative but more technical de-blur adjustments.
So it basically boils down to whether to do one's own output sharpening, which is the best practice in most cases, or leave it to a proven high quality treatment by a trusted service.
Cheers,
Bart
