Pages: [1] 2   Go Down

Author Topic: A focus magic sharpening sample (repost)  (Read 5525 times)

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
A focus magic sharpening sample (repost)
« on: November 28, 2013, 01:10:44 am »

Hi,

This image shows the effect of Focus Magic on an image shot with smallish aperture and not optimal focusing.

Distagon 40/4, f/16 - f/22, focus chosen for maxim depth of field.

Capture sharpening in LR (could be pushed a bit more)

Sharpening by Focus Magic as shown within framed areas, the rest is LR 5 capture sharpening alone.

Note how much detail FM extracts from rock surfaces. FM Rocks!

Note, this effect is a combination of capture sharpening in Lightroom and Focus Magic. Focus Magic it self will not give this kind of result. (My guess is that FM handles defocus very well and LR handles diffraction very well). Normally I would use radius 1.3 for f/22 in LR but this case 1.0 was used.

Best regards
Erik
« Last Edit: November 28, 2013, 01:13:06 am by ErikKaffehr »
Logged
Erik Kaffehr
 

Ligament

  • Full Member
  • ***
  • Offline Offline
  • Posts: 212
Re: A focus magic sharpening sample (repost)
« Reply #1 on: November 28, 2013, 01:52:10 am »

Thanks for the post. I love FM as well.

However, with the LR settings you have set, my understanding is LR is using 100% deconvolution sharpening.

You are then using another deconvolution pass when using FM.

It seems to me redundant and possibly detrimental to the image to do it like this. I could of course be wrong.

EDIT: I see you shot this at f22. With the severe diffraction at this f stop perhaps two rounds of deconvolution are indeed needed! I rarely shoot above f8 so have not needed to overcome such extreme diffraction personally. I have no problem with anybody using high f stops BTW, do what you need to get the job done.
« Last Edit: November 28, 2013, 01:55:51 am by Ligament »
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: A focus magic sharpening sample (repost)
« Reply #2 on: November 28, 2013, 02:08:33 am »

Hi,

I was somewhat surprised. Neither FM or the LR really did the job well enough, but the combo shines. I also tested Topaz InFocus but it made havoc of the three tops lower right.

Best regards
Erik

Thanks for the post. I love FM as well.

However, with the LR settings you have set, my understanding is LR is using 100% deconvolution sharpening.

You are then using another deconvolution pass when using FM.

It seems to me redundant and possibly detrimental to the image to do it like this. I could of course be wrong.

EDIT: I see you shot this at f22. With the severe diffraction at this f stop perhaps two rounds of deconvolution are indeed needed! I rarely shoot above f8 so have not needed to overcome such extreme diffraction personally. I have no problem with anybody using high f stops BTW, do what you need to get the job done.
Logged
Erik Kaffehr
 

Fine_Art

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1172
Re: A focus magic sharpening sample (repost)
« Reply #3 on: November 28, 2013, 12:25:52 pm »

A lot of that is fake detail. You can see it in the trees below the rocks, they take on a synthetic watercolor ripple effect. The only difference on the rocks is the fake texture looks somewhat compatible. You could put a layer of completely random "grain" over the rocks. Many people would see the changing contrasts as detail. It means nothing.

This is not the correct deconvolution.
Logged

Fine_Art

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1172
Re: A focus magic sharpening sample (repost)
« Reply #4 on: November 28, 2013, 01:50:24 pm »

This is a sample of deconvolution. 100% crop unsharpened. You can see mild softness.
Deconvolved you can see mild artifacting.

That app is not color controlled by the way so the export colors from raw therapee look different. Ignore the color, look at the detail.
« Last Edit: November 28, 2013, 01:53:30 pm by Fine_Art »
Logged

Fine_Art

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1172
Re: A focus magic sharpening sample (repost)
« Reply #5 on: November 28, 2013, 02:20:47 pm »

This is with Bart's supersampling method. The artifacts disappear. The impact is also reduced so you might want more cycles to taste if it is for a print. This level is appropriate for screen, mild artifacting is smeared away in a print.
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: A focus magic sharpening sample (repost)
« Reply #6 on: November 28, 2013, 02:41:40 pm »

Thanks for your observations, much appreciated! I might have been to enthusiastic about the results, you comments make a lot of sense!

Best regards
Erik


A lot of that is fake detail. You can see it in the trees below the rocks, they take on a synthetic watercolor ripple effect. The only difference on the rocks is the fake texture looks somewhat compatible. You could put a layer of completely random "grain" over the rocks. Many people would see the changing contrasts as detail. It means nothing.

This is not the correct deconvolution.
Logged
Erik Kaffehr
 

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: A focus magic sharpening sample (repost)
« Reply #7 on: November 28, 2013, 04:27:42 pm »

Thanks for your observations, much appreciated! I might have been to enthusiastic about the results, you comments make a lot of sense!

Best regards
Erik



Erik, your enthusiasm is understandable and appreciated at least by me because I just discovered my enthusiasm trying to sharpen that "fox" image in CS3's ACR and was bedazzled by what I could bring out in detail until I put it up against Fine_Art's last correction and became disappointed. Then I became even more bedazzled when I decided to top Fine_Arts sharpening by using LR4's sharpening and was even more amazed that it was better than Fine_Art's attempt.

THEN!...I was pissed beyond belief that the Fox image was untagged and that's when I knew I was dealing with amateurs here or possibly folks that need prescription glasses. I'm having to get by with 1.5x magnification reading glasses I got at Walmart.

But yeah, Erik, the mountain image is over sharpened with regards to micro detail which I doubt any human on the planet would be able see such detail whether they were viewing the actual scene or a print of it but I don't blame the software more than I question the level of quality of folk's eyesight here.
« Last Edit: November 28, 2013, 06:55:02 pm by Tim Lookingbill »
Logged

Fine_Art

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1172
Re: A focus magic sharpening sample (repost)
« Reply #8 on: November 28, 2013, 05:39:14 pm »

So what, in your mind, is the profound value of putting tags on a small crop out of an image, such that it is worth getting pissed beyond belief about?
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: A focus magic sharpening sample (repost)
« Reply #9 on: November 28, 2013, 06:47:16 pm »

So what, in your mind, is the profound value of putting tags on a small crop out of an image, such that it is worth getting pissed beyond belief about?

It puts your credibility as an authority on this subject in question (something I don't want to find out about after an editing session) because it points to the fact you aren't aware of the slight variances in display calibration/profile linearity between a wide range of display manufacturer's appearance of contrast/black point tonal roll off out on the market combined with the fact the shadow bump engineered into the sRGB 2.1 color space that we all convert to for web amplifies these nonlinear variances affecting the appearance of local contrast because some untagged sRGB images have applied by some CM browsers/OS systems the monitor's traditional 2.2 gamma curve.

This affects the appearance of sharpness due to a lack of or amplification of clarity in the shadow regions that might make what looks sharp on someone else's calibrated display look like lens diffraction on another's.

At least with a tagged sRGB image (we're now having to assume with the fox image) we've ruled out this particular synthetic color space's nonconventional 2.2 gamma affecting sharpness appearance caused by differences in contrast.

I had to apply some Clarity to the fox image to get it to look more sharp because global combined with local combined with micro contrast must transition together to work hand in glove to give the illusion of sharpness without amplifying edge halos. Now I'm not sure if the Clarity adjust was necessary due to this image not being tagged with its intended gamma curve.
« Last Edit: November 28, 2013, 06:49:44 pm by Tim Lookingbill »
Logged

Fine_Art

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1172
Re: A focus magic sharpening sample (repost)
« Reply #10 on: November 28, 2013, 09:43:23 pm »

It puts your credibility as an authority on this subject in question (something I don't want to find out about after an editing session) because it points to the fact you aren't aware of the slight variances in display calibration/profile linearity between a wide range of display manufacturer's appearance of contrast/black point tonal roll off out on the market combined with the fact the shadow bump engineered into the sRGB 2.1 color space that we all convert to for web amplifies these nonlinear variances affecting the appearance of local contrast because some untagged sRGB images have applied by some CM browsers/OS systems the monitor's traditional 2.2 gamma curve.

This affects the appearance of sharpness due to a lack of or amplification of clarity in the shadow regions that might make what looks sharp on someone else's calibrated display look like lens diffraction on another's.

At least with a tagged sRGB image (we're now having to assume with the fox image) we've ruled out this particular synthetic color space's nonconventional 2.2 gamma affecting sharpness appearance caused by differences in contrast.

I had to apply some Clarity to the fox image to get it to look more sharp because global combined with local combined with micro contrast must transition together to work hand in glove to give the illusion of sharpness without amplifying edge halos. Now I'm not sure if the Clarity adjust was necessary due to this image not being tagged with its intended gamma curve.

First of all I do not claim to be an authority on the subject. If I was writing something academic it would not be on a web forum.

Second you are talking about illusion which is USM halos. This thread is about deconvolution programs. You always seem to come out talking about PS or lightroom. In the sharpened images I posted there are fairly harsh jumps in luminance from deconvolution. Jumps that create an aliased look. You are looking for the ruts on the sides of edges created by USM.

Third it is intended as SRGB with Gamma 2.2. That is how I output from RT. Images Plus that I used does not know about color gamuts. It will display colors with an unknown approximation based on sRGB. The math that it does is precise. If another app opens the jpg it will be based on the original RT output color apart from one thing. Part of the normal sharpening process I use in the program is a routine called adaptive contrast. It increase saturation as well as luminance contrast. That is why I stated the program is not color controlled. You can consider that saturation +5 (approximately) from the RT output.

If you have a version that you feel looks better, post it.
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: A focus magic sharpening sample (repost)
« Reply #11 on: November 29, 2013, 03:51:54 pm »

First of all I do not claim to be an authority on the subject. If I was writing something academic it would not be on a web forum.

Second you are talking about illusion which is USM halos. This thread is about deconvolution programs. You always seem to come out talking about PS or lightroom. In the sharpened images I posted there are fairly harsh jumps in luminance from deconvolution. Jumps that create an aliased look. You are looking for the ruts on the sides of edges created by USM.

Third it is intended as SRGB with Gamma 2.2. That is how I output from RT. Images Plus that I used does not know about color gamuts. It will display colors with an unknown approximation based on sRGB. The math that it does is precise. If another app opens the jpg it will be based on the original RT output color apart from one thing. Part of the normal sharpening process I use in the program is a routine called adaptive contrast. It increase saturation as well as luminance contrast. That is why I stated the program is not color controlled. You can consider that saturation +5 (approximately) from the RT output.

If you have a version that you feel looks better, post it.

Which 2.2 gamma? sRGB 2.1 Microsoft, HP and Apple came up with that has the non-standard gamma curve? Or the 2.2 gamma curve that's in AdobeRGB? That's all I'm talking about. I didn't raise the question about color.

You'll have to excuse me for not correlating and seeing the benefit of all you outlined above concerning deconvolution from RT just from examining your posted sample. I didn't even know that that was what you were trying to communicate even though you sounded so sure that you were giving helpful information that could be used by contributors to this thread including Erik. I was trying to gleem and understand the benefits from your initial response until I noticed I was working on an untagged image. You didn't come across as academic in what you were trying to explain.

There's only one thing I'm concerned about with regards to sharpening and that is to get the best edge detail applied evenly across the entire image without artifacts. IOW achieving a natural look to a sharpened image. I'm not interested in micro sparkles showing up in order to get this. LR's PV2012 changed their Masking slider behavior that handles those types of artifacts quite well at least way much better than previous process versions.

I also notice in discussion of this kind that I find most folks don't know how to explore and use the tools they have to get the best results and start to blame the software instead of understanding how to create an illusion that is a sharpened image and that requires an intimate knowledge of what the tools are doing to the image on pixel level. Erik gave the perfect demonstration of this. I don't care what tools a photographer has to create the best looking image. I care more about them learning how to use those tools.

I know what a well sharpened image is suppose to look like down to the pixel viewing level because I know what that will look like when downsampled/upsampled for viewing on display and prints at the right distances and resolution and that's all that matters.
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: A focus magic sharpening sample (repost)
« Reply #12 on: November 29, 2013, 04:50:16 pm »

Here's my attempt at sharpening in LR4.4. I first assigned Epson sRGB (the same as sRGB2.1) and converted to sRGB2.1 in Photoshop and resaved at maximum quality just to confirm the space and not introduce more artifacts from jpeg recompression and then worked on the sharpening and tonality for the results below.
Logged

Fine_Art

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1172
Re: A focus magic sharpening sample (repost)
« Reply #13 on: November 29, 2013, 11:35:39 pm »

Here's my attempt at sharpening in LR4.4. I first assigned Epson sRGB (the same as sRGB2.1) and converted to sRGB2.1 in Photoshop and resaved at maximum quality just to confirm the space and not introduce more artifacts from jpeg recompression and then worked on the sharpening and tonality for the results below.

Not bad. It is harder contrast with halos. It's interesting that the halos have mosquito noise. You were starting from a jpg where as I am starting from tif.

For any level of crunchyness, I think it will always look better using a deconvolution method over a USM method. Halos are quite annoying.

« Last Edit: November 30, 2013, 12:03:08 am by Fine_Art »
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: A focus magic sharpening sample (repost)
« Reply #14 on: November 30, 2013, 01:56:25 am »

Hi,

I don't know what camera you used for the fox shot, but the images show color aliasing very similar to what I have seen in some of mine P45+ images. I have seen similar artifacts on one of my Sony cameras, the Alpha 55 SLT (16 MP APS-C). I guess that one had a weak OLP filter, or possibly none.

Regarding deconvolution, it often causes ringing artefacts, also it requires a known PSF (point spread function). A gaussian can be used for diffraction. Smart sharpen in Photoshop has two kernels, one for defocus and another one called "gaussian" that is more useful for diffraction.

Best regards
Erik

Not bad. It is harder contrast with halos. It's interesting that the halos have mosquito noise. You were starting from a jpg where as I am starting from tif.

For any level of crunchyness, I think it will always look better using a deconvolution method over a USM method. Halos are quite annoying.


Logged
Erik Kaffehr
 

Fine_Art

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1172
Re: A focus magic sharpening sample (repost)
« Reply #15 on: November 30, 2013, 12:08:33 pm »

Yes, the fox shots are the Sony A55v with the Minolta 300 f4G. The fox was in the wild. I was following him when he became interested in a crow that landed nearby. I was less than 20 ft away.

The magenta blooming on highlights is mostly from not taking off the protective filter which amplifies the old film design shortcomings of that lens on digital. I know you have mentioned you have the same issue with that lens before.
Logged

Fine_Art

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1172
Re: A focus magic sharpening sample (repost)
« Reply #16 on: November 30, 2013, 12:20:24 pm »

Hi,

I don't know what camera you used for the fox shot, but the images show color aliasing very similar to what I have seen in some of mine P45+ images. I have seen similar artifacts on one of my Sony cameras, the Alpha 55 SLT (16 MP APS-C). I guess that one had a weak OLP filter, or possibly none.

Regarding deconvolution, it often causes ringing artefacts, also it requires a known PSF (point spread function). A gaussian can be used for diffraction. Smart sharpen in Photoshop has two kernels, one for defocus and another one called "gaussian" that is more useful for diffraction.

Best regards
Erik


yes, controlling the ringing becomes the limit of the process. The PSF also changes all over the image so all our consumer level software is an approximation even if you take the time to test for a PSF. I used a gaussian 5x5 pixels for that. I also did a few very mild blurrings between groups of deconvolution cycles. That tends to control the artifact buildup. I left all artifacts in which you can see on the green background. The process was the same on the whole image limited by the adaptive settings. Of course a layer would be used for a real image. Topaz remask in their FX lab stand alone program is very good.
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: A focus magic sharpening sample (repost)
« Reply #17 on: November 30, 2013, 02:51:36 pm »

Not bad. It is harder contrast with halos. It's interesting that the halos have mosquito noise. You were starting from a jpg where as I am starting from tif.

For any level of crunchyness, I think it will always look better using a deconvolution method over a USM method. Halos are quite annoying.



I don't see any halos along hair edges in my attempt on the fox image. The what you call mosquito noise along edges of hair strands won't be seen in a print due to the fact stochastic dithering of inkjet resolution won't have enough dot resolution to render it.

On your last attempt I see a lot of spectral like white dots peppered throughout and an overall softness to the entire image which is why just focusing on edge detail does not make a sharp image on its own. Like I said before global, local and micro contrast must transition together to create the illusion of sharpness in an image, not just edge sharpness.

I still haven't seen in this thread what deconvolution looks like and how it directly benefits the appearance of sharpness.
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: A focus magic sharpening sample (repost)
« Reply #18 on: November 30, 2013, 04:49:10 pm »

This one?

http://www.luminous-landscape.com/forum/index.php?topic=45038.0

Best regards
Erik


I still haven't seen in this thread what deconvolution looks like and how it directly benefits the appearance of sharpness.
Logged
Erik Kaffehr
 

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: A focus magic sharpening sample (repost)
« Reply #19 on: November 30, 2013, 07:28:19 pm »

Just remembered the Roger Clark page that shows a demo of its effect on sharpness appearance.

Thanks for the reminder, Erik.
Logged
Pages: [1] 2   Go Up