I use ACR/LR to process to process most of my Nikon D800e raw files, but now I see that the Lens Correction tool in Capture One Pro 10 has been improved and now includes Lens Diffraction Compensation. This is a feature that compensates for the loss of sharpness and micro contrast that occurs in images shot using small apertures. I presume this is some type of deconvolution algorithm and wonder how it compares to tools that I already have with respect to diffraction mitigation.
Hi Bill,
'Diffraction correction', as implemented in Capture One 10, is a very useful addition and especially when viewed from a workflow perspective. While the control freak in me would have liked some manual control, it actually does pretty well on automatic. It uses deconvolution (without generating noticeable artifacts) and it needs some EXIF metadata (like aperture number) to do its thing
on Raw files. That last bit about Raw is important because it becomes a part of the parametric input for the final rendering. So we have Lateral Chromatic Aberration correction
and diffraction deconvolution to improve the rendering of micro-contrast/color. The workflow benefit is in that shots taken with different apertures will look like having almost identical micro-contrast. So shooting scenarios are more liberated from picking a fixed aperture, maybe one a bit wider than preferred, in order to optimize the rendering of detail. Restraint no more, use what is needed for DOF instead.
Unfortunately it is not perfect, but then what is? While it works very good at different shooting distances, it might do fractionally better if the user could adjust for shooting distance. At closer focusing distances the magnification factor increases, and so does the size of the diffraction pattern. So in a perfect world, one would compensate for shorter shooting distances by increasing the radius of the Airy disc pattern PSF. However, as said, the automatic correction does work quite well both on close-up shots as well as on distant subjects. So apparently, their algorithm does pick up some of the required settings from context, rather than from metadata or user input.
The fact that it might be slightly under-correcting, and that it doesn't address other sources of lens/capture-blur, leaves some room for post-processing deconvolution. A cascade of deconvolutions is mathematically equal to a single deconvolution with a larger radius. In addition, we have the opportunity to mix the deconvolution using a diffraction pattern PSF with a subsequent deconvolution that uses a more Gaussian PSF. Best of both worlds?
As a compromise, instead of a Gaussian PSF based deconvolution as second stage, we can already get quite far with USM based sharpening when diffraction blur is mostly absent. For that, the 'Halo suppression' that is now added to the sharpening controls is a very useful addition. Moreover, sharpening with different settings (not only amount) can also be applied in Adjustment layers so that local sharpening adjustments (if need be in multiple layers with different settings for different areas of the image) can be achieved instead of, or in addition to, generic sharpening (which can employ sharpness fall-off sharpening).
LR/ACR have some type of relatively weak deconvolution in LR/ACR by moving the detail slider to the right and we have smart sharpen in Photoshop which uses deconvolution. Deconvolution plugins include Focus Magic, Topaz InFocus and Topaz Detail. Images Plus has good adaptive Richardson-Lucy and van Cittert deconvolution algorithms, although this involves the use of a separate program. Finally, RawTherapee has Richardson-Lucy and excellent demosaicing.
Yes, we have useful tools, but some of them are
needed to compensate for shortcomings in the earlier stages of capturing and rendering our Raw images. If we tackle some of the issues earlier, then we need to do less work later on. A program like Topaz Detail is a bit different because it addresses the modulation of different feature sizes, also those that do not necessarily count as micro-detail AKA sharpening.
Does C! add to what I already have?
In my opinion, yes. However, excellent tools like FocusMagic and similar come a long way in restoring some of the data but they do require additional postprocessing steps.
It is now possible to create quite well-sharpened output, e.g. for Web publishing, straight from the Raw converter. The output sharpening proofing options in C1 version 10 are another very useful addition for a faster workflow. By activating multiple Process recipes at the same time, C1 is able to render several different versions of a single Raw conversion in one click, e.g. 1 master rendering at full size for further retouching, 1 version for Web publishing, and 1 version for printing at a specific size and PPI. While the multiple outputs were already possible in earlier versions, the fact that they are each individually better sharpened makes it a much more useful feature.
Cheers,
Bart
