Pages: 1 2 3 [4]   Go Down

Author Topic: Which Capture Stage Sharpeners?  (Read 27648 times)

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8914
Re: Which Capture Stage Sharpeners?
« Reply #60 on: March 10, 2015, 09:28:43 am »

In this video (start 6:00), The Topaz rep uses the Edge Softness slider to correct an artifact which he calls "ringing." Is this artifact the same kind you refered to as "...ringing artifacts, repeated edges of gradually lower amplitude"? If yes, why does he recommend the Edge Softness slider to fix it while you say to use the Supress Artifacts slider?

Yes, but the ringing is in a sense caused by the overly sharp edges. So the Edge Softness control takes away the cause for the ringing, but it doesn't address the ringing artifacts themselves. The Suppress Artifacts control does that, after the artifacts were already created. So the Softness control prevents sharp edges that could cause artifacts, and Suppress Artifacts control reduces the artifacts that were already created.

Quote
It would help if I saw exactly the types of artifacts each of these sliders are designed to mitigate. How does  "Ringing" look different from "stair stepping?"

Ringing looks like waves with repeating periodic brighter and darker patterns around high contrast edges and lines. This document describes a method to detect and remove the artifacts, the images will show you some examples of ringing artifacts. Stair stepping turns a straight angled edge into a stepped (like stairs) edge with straight edge segments instead of a smooth continuous edge or line.

When you have trouble getting the correct settings in InFocus, start with the Sharpen panel settings maxed out for Micro contrast and Sharpness, and a Sharpness Radius setting of say 0.7 or 0.8 . Also set the Edge Softness and Suppress Artifacts sliders to their minimum. Now start with increasing the deconvolution method's radius slider while carefully avoiding double edges and ringing and stairstepping artifacts from appearing. The moment the artifacts are starting to show, use the Edge Softness control to reduce them again if in Estimate mode, and or use the Suppress Artifacts mode to take away the onset of the artifact generation. When in doubt, reduce the radius.

Only then adjust the Sharpness panel controls, sharpness radius first (usually smaller than the deconvolution radius), sharpness and micro contrast next, until you get an acceptable artifact free result. This way you are guaranteed artifact free results that will stand large format output without magnified artifacts.

Don't be surprised if your deconvolution radius settings turn out lower than they used to, because you are now no longer over-sharpening by using a too large radius. Also remember that InFocus is a Capture sharpening tool, not for Creative 'sharpening', that's where their 'Detail' plugin comes in.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

texshooter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: Which Capture Stage Sharpeners?
« Reply #61 on: March 10, 2015, 12:47:00 pm »

Only then adjust the Sharpness panel controls, sharpness radius first (usually smaller than the deconvolution radius), sharpness and micro contrast next, until you get an acceptable artifact free result. This way you are guaranteed artifact free results that will stand large format output without magnified artifacts.

A Topaz official said InFocus' Sharpness panel uses USM technology (albeit a new and improved version of USM). (Ref. video 12:30-13:30)
https://m.youtube.com/watch?v=zChtYpEBFBQ

So I'm surprised to hear you suggest the  Sharpness panel during the capture sharpening phase. Shouldn't I turn off InFocus' Sharpness panel (after running the estimate deblur command) and postpone USM methods for the end of the entire edit workflow? The part you said about maxing out the  Sharpness panel sliders as a temporary visual aid only to help estimate the deconvolution radius...I understand that however....and will give it a try, thanks.
« Last Edit: March 10, 2015, 01:28:49 pm by texshooter »
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8914
Re: Which Capture Stage Sharpeners?
« Reply #62 on: March 10, 2015, 01:39:32 pm »

A Topaz official said InFocus' Sharpness panel uses USM technology (albeit a new and improved version of USM). (Ref. video 12:00-13:30)
https://m.youtube.com/watch?v=zChtYpEBFBQ

Yes, an "advanced type of USM technology". That means that is tends to enhance edge detail relatively more than subtle detail, that's why they have also added a Micro Contrast control to fill in that part of the equation. But what these tools do, is on top of the deconvolved image information, so that's fine as long as we don't exaggerate it. After all, we are at the stage of Capture sharpening. That's also why it's important to not generate deconvolution artifacts, they will only get worse.

Quote
So I'm surprised to hear you suggest the  Sharpness panel during the capture sharpening phase. Shouldn't I turn off InFocus' Sharpness panel (after completing capture sharpening) and save USM methods for the end of the entire edit workflow? The part about maxing out the  Sharpness panel sliders as a starting fulcrum to help estimate the deconvolution radius...I understand that however....and will give it a try, thanks.

The two work together as a cascade, but as I suggested the sharpening panel can also be used to temporarily exaggerate the effect of the deconvolutions, as to get a better view of when things start getting out of hand there.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 798
    • Hikes -more than strolls- with my dog
Re: Which Capture Stage Sharpeners?
« Reply #63 on: March 10, 2015, 05:57:02 pm »

... would be even more useful if one could import one's own PSF...

You know who I think is taking advantage of that approach these days is Olympus, with the EM5II's custom raw conversion PS plug-in for the 64MP High Resolution images.  Taking a look at the resulting MTF curve, it appears to me that they use a sensor specific PSF, pushing it to taking advantage of the fact that supersampling leaves the raw curve quite a ways from Nyquist OOC.

Jack
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8914
Re: Which Capture Stage Sharpeners?
« Reply #64 on: March 11, 2015, 06:05:07 am »

You know who I think is taking advantage of that approach these days is Olympus, with the EM5II's custom raw conversion PS plug-in for the 64MP High Resolution images.  Taking a look at the resulting MTF curve, it appears to me that they use a sensor specific PSF, pushing it to taking advantage of the fact that supersampling leaves the raw curve quite a ways from Nyquist OOC.

Hi Jack,

It might well be that Olympus uses some form of Point Spread Function (PSF) based adjustment  during the Raw conversion. As far as the combined exposure PSF from supersampling alone goes, that would be rather predictable and highly reproducible. It's the convolution of the PSFs of the subexposure positions. Some of the sub-exposures have overlapping sampling areas, so one could envision a PSF from a (up to) 200% fill-factor sample per new super-sampling grid. The individual samples probably have a Gaussian PSF shape, just like common sensels apparently do.

My PSF-generator tool offers the possibility to define a 200% fill-factor, which flattens (lowers kurtosis of) the Gaussian PSF shape a bit more than a point sampler would. Since Gaussian shaped PSFs have long tails, the individual PSFs would have to use a relatively large kernel support, so it may be more efficient to use a single cascaded PSF from the overlapping PSFs after R/G/B reconstruction, but a lot depends on the details which approach is best. When still dealing with sensor DNs (ADUs), one can use fast binary shifts instead of floating point math multiplications, that's what the option to generate binary shifted type of kernel numbers is for.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==
Pages: 1 2 3 [4]   Go Up