Pages: 1 [2] 3 4   Go Down

Author Topic: Which Capture Stage Sharpeners?  (Read 27632 times)

texshooter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: Which Capture Stage Sharpeners?
« Reply #20 on: March 02, 2015, 04:33:18 am »



I get the impression that that is what happens in Canon's Digital Photo Professional (DPP) raw converter, when the Digital Lens Optimizer (DLO) is used. The Raw data is recalculated to a version without lens aberrations, with lens blur removed, and resaved in the Raw file container as an additional raw dataset, thus almost doubling the file size. That optically corrected raw image data is then demosaiced into a much better quality, and less sharpening is required as a post-processing step.

Incidentally, Canon recommends all sharpening be turned off before starting the DLO function....

http://web.canon.jp/imaging/dlo/howto/index.html

Which would seem to confirm your hunch. Funny thing ACR lacks this sophistication. I'm disappointed.

Logged

Jimbo57

  • Full Member
  • ***
  • Offline Offline
  • Posts: 180
Re: Which Capture Stage Sharpeners?
« Reply #21 on: March 02, 2015, 04:40:18 am »

I am another of those who finds that Lightroom handles the capture sharpening of my D800E and D810 Raws to my satisfaction.

Similarly, as I always print from Lightroom, I am happy with the output sharpening it applies.

If, for creative reasons, I want to sharpen or soften an image, either in the whole or selectively, once more Lightroom seems to meet my needs.

When I use CS6 or any of the Nik or Topaz programs, I use them from within Lightroom and rarely fiddle with the sharpening in them unless for extreme creative purposes.

Just an observation - I now hear judges comment more about "over-sharpened" images than about "soft" images.
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Which Capture Stage Sharpeners?
« Reply #22 on: March 02, 2015, 06:39:13 am »

Incidentally, Canon recommends all sharpening be turned off before starting the DLO function....

Yes, but not because it interferes with the sequence of events, but to better judge what DLO does on its own ("In order to best understand the effect of the Digital Lens Optimizer"). DLO allows to set the strength of the optical correction, between 0 and 100, with a default at 50, and it would become harder to judge the effect if additional sharpening (applied after demosaicing) is already included in the preview. It's better to correct before, rather than after conversion, but one can still apply additional sharpening after conversion.

Quote
Funny thing ACR lacks this sophistication. I'm disappointed.

Well, Canon of course knows the details of their Raw files, and lens data, better than those who have to reverse engineer some of it. If Adobe would use better deconvolution algorithms for sharpening (and allow to apply them spatially variant, across the image), the gap would already close somewhat. I suppose that for speed reasons alone they will not attempt to fully correct images, although a converter like Capture One does offer a simple sharpness fall-off correction, which is not that bad a compromise at all.

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

t6b9p

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 70
    • BeyondVisible
Re: Which Capture Stage Sharpeners?
« Reply #23 on: March 02, 2015, 11:57:06 am »

Quote
It would make sense to even do most of it before demosaicing, because that would produce more accurate data to demosaic. However, that requires much more complicated Raw conversion.

I would be interested to know what steps/software a "complicated Raw conversion" would require in order to implement deconvolution prior to demosaicing.

I am shooting 830nm IR which behaves similarly to a monochromatic sensor as the dyes in the Bayer array are basically transparent to IR at this wavelength. Therefore I would like to perform deconvolution sharpening to the image prior to demosaicing and then move on to CS for other edits.
Logged

texshooter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: Which Capture Stage Sharpeners?
« Reply #24 on: March 02, 2015, 12:32:31 pm »

I would be interested to know what steps/software a "complicated Raw conversion" would require in order to implement deconvolution prior to demosaicing.

I think Bart said that Canon's DPP does some deconvolution prior to demosaicing, not via the sharpening slider per se, but rather via the DLO function. In other words, the DLO function removes some of the hardware-induced blur by means of a deconvolution process. As far as capture sharpening goes (using the sharpen slider in DPP), I would guess demosaicing is done first. Not sure.

But if you are converting infrared RAWs, you're biggest problem is not deconvolution or demosaicing, but rather white balance. The web is replete with advice that Adobe's ACR does a lousy job interpreting IR RAW files. I would use Canon's DPP (if you should Canon) for my IR conversion needs, not ACR or Lightroom. For those interested in IR RAW conversion techniques...

http://www.lifepixel.com/tutorials/infrared-photoshop-videos

« Last Edit: March 02, 2015, 01:12:05 pm by texshooter »
Logged

t6b9p

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 70
    • BeyondVisible
Re: Which Capture Stage Sharpeners?
« Reply #25 on: March 02, 2015, 12:38:41 pm »

I should have specified that I shoot Nikon and with regards to WB - it is not a problem - I have been using UniWb for 10+ years.
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Which Capture Stage Sharpeners?
« Reply #26 on: March 02, 2015, 12:41:30 pm »

I would be interested to know what steps/software a "complicated Raw conversion" would require in order to implement deconvolution prior to demosaicing.

I am shooting 830nm IR which behaves similarly to a monochromatic sensor as the dyes in the Bayer array are basically transparent to IR at this wavelength. Therefore I would like to perform deconvolution sharpening to the image prior to demosaicing and then move on to CS for other edits.

Hi, that's right, the Bayer CFA is rather transparent to IR (although 830nm isn't that extreme), as many dyes are.

To pre-process the file, e.g. to address diffraction (which is more pronounced at 830nm) one would need to remap the sensels for residual lens distortion, and deconvolve for diffraction (if focused accurately, otherwise defocus is another thing to take care of).

If distortion is the issue, then one virtually re-positions the recorded positions to their supposed origins before distortion, and does the Demosaicing based on the interpolated current positions on the imaginary original. One then has a new re-sampled de-warped CFA, which could be deconvolved to restore original input resolution and resampling losses, and then demosaic it on that interpolated/restored CFA pattern.

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

t6b9p

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 70
    • BeyondVisible
Re: Which Capture Stage Sharpeners?
« Reply #27 on: March 02, 2015, 01:07:49 pm »

Thanks Bart

I already understand what needs to be done, I was more interested in what software options were available to implement those steps e.g. Dcraw, Rawtherapee etc. In other words I was looking for a good solution to implement FocusMagic deconvolution prior to passing off to CS

Also, at 830nm, the Bayer dyes are not quite equally transparent so require slight scaling, followed by subsequent processing without demosaicing (after all, if the sensor is responding like a monochrome sensor why treat it like a Bayer sensor and bother demosaicing).
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Which Capture Stage Sharpeners?
« Reply #28 on: March 02, 2015, 03:27:45 pm »

I already understand what needs to be done, I was more interested in what software options were available to implement those steps e.g. Dcraw, Rawtherapee etc. In other words I was looking for a good solution to implement FocusMagic deconvolution prior to passing off to CS

Okay. Depending on what needs to be addressed, I'm thinking in the direction of the cylindrical resampling that ImageMagick allows, AKA Elliptical Weighted Averaging (EWA), because that allows to resample at various angles with similar quality in all directions, which is useful for distortions.

Quote
Also, at 830nm, the Bayer dyes are not quite equally transparent so require slight scaling, followed by subsequent processing without demosaicing (after all, if the sensor is responding like a monochrome sensor why treat it like a Bayer sensor and bother demosaicing).

Yes, so after minor filter transparency scaling one could skip the warping and demosaicing, and instead just use deconvolution to restore from the diffraction. DCraw should allow to produce the first unmodified output, with the -D parameter, in linear gamma 16-b/ch output with the -4 parameter, and to TIFF with the -T parameter. Instead of the -D parameter, you could play with the -d parameter, and add a channel weighting with the -r 1 1 1 1 parameter (with the weights not set to 1 for modifying the per channel transparency). You may even get lucky with the -a Auto white-balancing parameter. Then take the result through FocusMagic or similar, and introduce an output gamma for display (or leave it in linear gamma for quantitative measurements).

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

t6b9p

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 70
    • BeyondVisible
Re: Which Capture Stage Sharpeners?
« Reply #29 on: March 02, 2015, 03:54:29 pm »

thanks again.

I have tried the Dcraw route before for unmodified output, as well as scaling the RGB channels, (many years ago so may retry just to reconfirm) and found it to be mediocre but will check out the ImageMagik suggestion.
Logged

dwswager

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1375
Re: Which Capture Stage Sharpeners?
« Reply #30 on: March 05, 2015, 03:16:03 pm »

I hesitate to admit it, but I use Lightroom for the majority of my non-critical capture sharpening. My camera (Nikon D800e) lacks a blur filter and needs less capture sharpening than cameras with such a filter and the parametric workflow is a great convenience. For those images that will need to be brought into Photoshop and saved as TIFFs, I do use Focus Magic or the Topaz sharpeners. One problem for me with Topaz is that there is a bewildering number of adjustments in each application and is is not clear which of the Topaz tools would be best for a given image and what the optimal parameters might be (I lack Bart's sophistication). Focus Magic is much simpler to use.

With LR/ACR one can make use of a simplified deconvolution tool by moving the detail slider to the right and my testing shows that the results are not that much inferior to those obtained with Focus Magic. I respect Mark Segal's opinion, but PhotoKit is old technology for capture sharpening. However, it is good for creative and output sharpening.

Bill

I'm certainly no expert, but I do almost all capture sharpening in ACR, unless I just can't get it worked out to my satisfaction. Then I try Topaz InFocus.  I like being able to go back and forth with sharpening/noise simultaneously.    Have not used Focus Magic or PKS.

Infocus is supposed to be the Topaz tool of choice for capture sharpening while Detail is supposed to be the creative/output tool of choice.  But yeah, Topaz has too many tools that are similar to try and keep track!
Logged

texshooter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: Which Capture Stage Sharpeners?
« Reply #31 on: March 05, 2015, 03:27:54 pm »


Infocus is supposed to be the Topaz tool of choice for capture sharpening while Detail is supposed to be the creative/output tool of choice.  But yeah, Topaz has too many tools that are similar to try and keep track!

I use Infocus for all my capture sharpening because I like how it estimates the de-blur for me. With ACR, I don't know what I'm getting really. Although, I'm sure experts like Jeff Schewe, who wrote the bible on the subject, will probably vouch for deconvolving with ACR's sliders, instead. Except Bart, of course.
Logged

dwswager

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1375
Re: Which Capture Stage Sharpeners?
« Reply #32 on: March 05, 2015, 04:27:35 pm »

I use Infocus for all my capture sharpening because I like how it estimates the de-blur for me. With ACR, I don't know what I'm getting really. Although, I'm sure experts like Jeff Schewe, who wrote the bible on the subject, will probably vouch for deconvolving with ACR's sliders, instead. Except Bart, of course.

If you hold the ALT (Windows) or the equivalent MAC key down in ACR/LR while moving the sliders it will show you what the Amount, Radius, Detail and Masking controls are doing.  Basically, I start with a high value on the amount and then hit the masking slider such that masks sharpening the sky or other areas like that.  Then I work the detail and radius sliders to suit and then back down the amount to what I think works best remembering that this is capture sharpening.
Logged

texshooter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: Which Capture Stage Sharpeners?
« Reply #33 on: March 06, 2015, 09:09:30 pm »

Basically, I start with a high value on the amount and then hit the masking slider such that masks sharpening the sky or other areas like that.  Then I work the detail and radius sliders to suit and then back down the amount to what I think works best remembering that this is capture sharpening.

Why would you need to mask the sky for capture sharpening? Assuming you are starting with a low noise image, if you need to mask the sky then your sharpening amount is too strong from the outset. True capture sharpening does not over-sharpen any area, skies or else.  Perhaps your intent is to combine creative and capture sharpening in a single pass (which is ok is you're short on time?  The way I see it, capture sharpening is the first pass of sharpening to be applied globally, no masking slider required. Afterwards, If I want the higher frequency details punched up a bit I will apply some creative sharpening using Smart Sharpen, Blend Ifs, sky masking, etc, while inside PS and after all editing is done.
« Last Edit: March 06, 2015, 09:15:24 pm by texshooter »
Logged

dwswager

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1375
Re: Which Capture Stage Sharpeners?
« Reply #34 on: March 06, 2015, 10:11:29 pm »

Why would you need to mask the sky for capture sharpening? Assuming you are starting with a low noise image, if you need to mask the sky then your sharpening amount is too strong from the outset. True capture sharpening does not over-sharpen any area, skies or else.  Perhaps your intent is to combine creative and capture sharpening in a single pass (which is ok is you're short on time?  The way I see it, capture sharpening is the first pass of sharpening to be applied globally, no masking slider required. Afterwards, If I want the higher frequency details punched up a bit I will apply some creative sharpening using Smart Sharpen, Blend Ifs, sky masking, etc, while inside PS and after all editing is done.

While a definite on night sports images where there is always high iso noise, but I also will mask in most images because I never really want sharpening is continuous tone areas, especially a clear sky.  Almost forgot, I also use masking to limit the effect on skin, especially in faces.
« Last Edit: March 06, 2015, 10:23:17 pm by dwswager »
Logged

Schewe

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6229
    • http:www.schewephoto.com
Re: Which Capture Stage Sharpeners?
« Reply #35 on: March 06, 2015, 10:51:08 pm »

Why would you need to mask the sky for capture sharpening?

Because capture sharpening should sharpen edges and not surfaces...

It's the edges where you want to concentrate sharpening and the edge mask in ACR/LR is designed to allow applying sharpening to edges and not apply sharpening to the non-edge (surfaces) in an image.

Large areas of color or tone without edges get no benefit from sharpening and even low ISO captures can have noise increased in in large open areas like skies.
Logged

texshooter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: Which Capture Stage Sharpeners?
« Reply #36 on: March 07, 2015, 12:03:24 am »

Because capture sharpening should sharpen edges and not surfaces...

This Youtuber uses a contrarian method of capture sharpening  in Camera Raw. He works from the bottom up, instead of top down. Is this a good idea?
One thing I noticed, though.   His method requires swinging the Detail slider far left which unpropitiously contravenes deconvolution.


« Last Edit: March 07, 2015, 12:06:04 am by texshooter »
Logged

Schewe

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6229
    • http:www.schewephoto.com
Re: Which Capture Stage Sharpeners?
« Reply #37 on: March 07, 2015, 01:01:41 am »

Is this a good idea?

No..this guy is a putz with an accent...you really (really) don't want to pay attention to this guy...while some of what he says is useful the problem is you won't know what part is useful.

In actuality, the first slider to select should be radius and is based on the edge frequency of the image....then amount then detail and finally, masking.

But hey, this is coming from somebody who has 1/2 a clue how to use ACR/LR capture sharpening....note, I did actually work with the engineers to help develop ACR/LR's capture sharpening.
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Which Capture Stage Sharpeners?
« Reply #38 on: March 07, 2015, 09:32:51 am »

In actuality, the first slider to select should be radius ...

I agree.

... and is based on the edge frequency of the image....

I disagree.

... then amount then detail and finally, masking.

I agree with Amount and Masking coming last.

Let's address the part I disagree with.

Capture sharpening is supposed to reverse the blurring effects that caused our original Capture lose some of the original sharpness/detail. This has nothing to do with the original subject's detail, but only with the Capture process induced blur. The subject is not to blame that it now blurred, it's the Capture process itself. If we are able to reverse the capture blur, the subject will be restored to it's original full 'unblurredness' (how's that for a scrabble word ;) ). Some (but not all) detail is already lost beyond recovery, so it won't become perfect, but we can still restore a lot.

While it's true that edge detail is important in relaying the impression of sharpness (after all our eyes do something similar), surface structure is very much part of how we perceive materials. It allows us to e.g. discriminate between cardboard and leather. It's therefore paramount to also Capture sharpen this properly. Obviously, we do want to reduce (or at least not amplify) some photon shot noise in smooth gradients like sky or other structureless smooth gradients.

The photon shot noise is added to our image of the scene by using a relatively short exposure, and will be amplified by the Capture and demosaicing process. That, and some other sources of noise, is not what we want to emphasize in our images, so we also need to address that as part of the Capture sharpening process. We want to improve the signal (detail) to noise ratio.

The part that the video tutorial got right, is that the order of the sharpening controls in the detail panel is not in an optimal sequence, but then the sharpening controls themselves are not optimal either.

Quote
But hey, this is coming from somebody who has 1/2 a clue how to use ACR/LR capture sharpening....note, I did actually work with the engineers to help develop ACR/LR's capture sharpening.

Sorry but I'm not overly impressed with the result, although the masking is useful to cover up some of the sharpening shortcomings.

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

dwswager

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1375
Re: Which Capture Stage Sharpeners?
« Reply #39 on: March 07, 2015, 10:29:04 am »

No..this guy is a putz with an accent...you really (really) don't want to pay attention to this guy...while some of what he says is useful the problem is you won't know what part is useful.

In actuality, the first slider to select should be radius and is based on the edge frequency of the image....then amount then detail and finally, masking.

But hey, this is coming from somebody who has 1/2 a clue how to use ACR/LR capture sharpening....note, I did actually work with the engineers to help develop ACR/LR's capture sharpening.

Good to know!  I read a long time ago that the order in the panel was the correct order to work, but have never found that totally satisfactory.
Logged
Pages: 1 [2] 3 4   Go Up