Ad
Ad
Ad
Pages: « 1 ... 5 6 [7] 8 »   Bottom of Page
Print
Author Topic: Dan Margulis Sharpening Action  (Read 103742 times)
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 6767


WWW
« Reply #120 on: October 08, 2007, 01:00:24 PM »
ReplyReply

Quote
Interesting (FAR more interesting than Dan's BTW)...

The old "slightly crunchy" is simply way too little (and often wrongly applied) because a low resolution display can't give you the proper feedback for making visual determinations. .............. The ONLY thing a computer display is good at showing you is color and dynamic range and detail for the display–if the display is the final image output. If the final output is a print, a display is a poor predictor of what you must do to the image.
[a href=\"index.php?act=findpost&pid=144646\"][{POST_SNAPBACK}][/a]

Jeff,

I've heard this often enough to be mindful of it, and more than that, to work around it. Firstly, I notice you qualify the comment with the term "low resolution display". At what resolution does a display cease to be "low resolution". I ask this, because my 1600*1200 LaCie 321, while not the highest resolution out there these days is not the lowest either, and I can see sharpening impacts quite clearly as discussed just below.

Secondly, I agree with you that generally speaking the result of sharpening on a display will look nastier than it comes out on paper. But that also depends partly on the magnification. By making a few benchmark tests to get an impression of the outer limits of on-display sharpening appearance at 50% magnification relative to the resulting print, I "kind of know" how far I can take the output sharpening before I can expect an overly crunchy printed result. I'm not saying its fool-proof, but more often than not it's reliable.

Mark
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
Schewe
Sr. Member
****
Offline Offline

Posts: 5255


WWW
« Reply #121 on: October 08, 2007, 03:18:12 PM »
ReplyReply

Quote
At what resolution does a display cease to be "low resolution". I ask this, because my 1600*1200 LaCie 321, while not the highest resolution out there these days is not the lowest either, and I can see sharpening impacts quite clearly as discussed just below.
[a href=\"index.php?act=findpost&pid=144666\"][{POST_SNAPBACK}][/a]

You might be able so see some of the results, yes...but not in the context of the final printed size. My 30" Cinemas display at about 104PPI. My iPhone has a 160PPI display (which is really nice and sharp but small). But even at 104PPI, that's only 1/3 (roughly) of an image at 300PPI. The ONLY way to to "see" the effective resolution of 300PPI on the 104PPI display is fiddle with the size of the image (screen zoom in Photoshop). But the fact is, there's simply no way to accurately gauge the real 300+PPI resolution of a print file when the display is, at best, 1/3 the image resolution.

I've heard of some possible 200PPI displays in the future, which would be better (more like 2/3 the resolution of the image) but that is a far cry if you work at 480PPI because it would still be less than 1/2. And, in the case of the Cinema at 104PPI, that's less that 1/4th the original at 480PPI.

This really brings up the very real difficulty inherent in sharpening by eye, because nothing that you see on the display is really a reliable gauge to the print when it comes to detail...
Logged
Brian Gilkes
Sr. Member
****
Offline Offline

Posts: 431


WWW
« Reply #122 on: October 08, 2007, 05:49:52 PM »
ReplyReply

And paper surface, printer resolution, dithering algorithms, local contrast, local colour combinations, lighting level , spectal distribution, tonal energy balance , composition directed  perception and viewer hangover all affect perceived detail.
Sometimes there is a bit much emphasis on sharpening routines.
In any  case, it's not  so much about  the hammer, but where to hit and how hard.
Cheers,
Brian
www.pharoseditions.com.au
PS There are some  crappy hammers around.
Logged
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 6767


WWW
« Reply #123 on: October 08, 2007, 05:59:21 PM »
ReplyReply

Jeff, you're right that it's far from exact, say compared with soft-proofing which IS quite accurate (save for the inevitable issue of direct versus reflected light); the point I'm making is that to some extent one can deal with this circumstance by making a few tests for the outer limits (display appearance versus prints), thereby providing useful - albeit subjective - guidance about "boundary conditions" - i.e. the range beyond which the print runs the risk of looking brittle.

In my several years of using PKS, for example, as long as I select the options consistently with the input, frequency and output conditions of the image, more often than not one simply doesn't need to exercise this judgment; however there are those images where one needs to watch this - and I can tell you exactly the most likely candidates: images that have been both captured sharpened and had a dose of "clarity" (in CR/LR), or "mid-tone contrast enhancement" or 20-50-0 USM in Photoshop. In these conditions, by the time one gets to output sharpening, it's possible for the image to become too brittle (again a matter of taste). To assess the likelihood of this result coming out of the printer, these benchmarks can be helpful. But it is subjective and it takes experience to know what one is looking at on the display as a predictor of the printed product.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #124 on: October 09, 2007, 02:48:40 PM »
ReplyReply

Having just acquired PK Sharpener, I'm undertaking the project of comparing it to my current workflow (Focus Magic + my midtone sharpening action). Since PKS has a plethora of options and I've only had it a short time, take the results with a small grain of salt, since I'm not nearly as intimately familiar with it as I am with my current workflow. If anyone more familiar with PKS notices any glaring errors in the settings, feel free to point them out.

The test image is crop from a 1Ds capture of downtown Trier, taken on an overcast day. DR is just barely within the limits of the sensor, and there is a lot of fine detail and a bit of noise to make getting the detail without too much noise a bit of a challenge. No noise reduction or sharpening was applied in ACR or afterward.

This post's comparison is limited to capture sharpening; PKS vs Focus Magic. PKS settings were as follows: Expert High Resolution Digital Capture, Medium Width. Focus Magic's settings were: 1 pass 2 pixel radius, 25%, then 1 pass 1 pixel radius, 25% source set to Digital Camera for both passes.

[attachment=3521:attachment]

Some points of interest:

In the bottom left area, the Merkur Spielothek sign is brought into clearer focus by FM than PKS, without the noise in the wall above the sign being accentuated as it is with PKS. In general, PKS accentuates noise much more than FM, but focuses sharp edges less well. When the PKS smoothing layer is turned on, the noise accentuation goes away, but edge sharpening becomes even less effective. PKS appears to be USM-based, whereas FM actually attempts to mathematically unravel the effects of blur. FM appears to me to do a significantly better job overall of undoing the effect of an AA filter than PKS and USM-based sharpening in general, which is why I don't even use my own action for capture sharpening.
« Last Edit: October 09, 2007, 03:02:09 PM by Jonathan Wienke » Logged

Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 6767


WWW
« Reply #125 on: October 09, 2007, 04:02:21 PM »
ReplyReply

Quote
Some points of interest:

In the bottom left area, the Merkur Spielothek sign is brought into clearer focus by FM than PKS, without the noise in the wall above the sign being accentuated as it is with PKS. In general, PKS accentuates noise much more than FM, but focuses sharp edges less well. When the PKS smoothing layer is turned on, the noise accentuation goes away, but edge sharpening becomes even less effective. PKS appears to be USM-based, whereas FM actually attempts to mathematically unravel the effects of blur. FM appears to me to do a significantly better job overall of undoing the effect of an AA filter than PKS and USM-based sharpening in general, which is why I don't even use my own action for capture sharpening.
[a href=\"index.php?act=findpost&pid=144885\"][{POST_SNAPBACK}][/a]

Jonathan, I use a 1Ds also and take many pictures like this, so I can say something about it - hopefully not too dumb. Firstly, I know that FM is using a deconvolution approach which is fundamentally different from what PKS, NIK and others do. But focusing on the results (pun intended), I'm viewing your comparison on a LaCie 321 display at 200% magnification. I find the edge detail on the roof of the building to the right of Gergen better with PKS. The building facades look a bit crisper and cleaner in the FM version, where the word "Merkur" is slightly more readable.

Three issues that are more important: (1) Your capture sharpen settings: try it again with Super Fine Edge sharpen and see what happens. I find often this setting can produce a cleaner result even if the image is not of the highest frequency. (2) The sharpening process is not complete. Capture sharpening is just the cake half or one-third baked. You need to do the Output Sharpening, and perhaps some Creative Sharpening targeted at any specfific issues. (3) Judging this on a display, as Jeff has pointed out, is not very reliable, but perhaps comparative reliability is not so bad. To overcome this issue for sure, the prints need to be compared. Of course that doesn't help us, but you would see better the differences and can describe them to us.

Hope this helps.

Mark
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
laughfta
Jr. Member
**
Offline Offline

Posts: 89


« Reply #126 on: October 09, 2007, 04:28:01 PM »
ReplyReply

Quote
The sharpening process is not complete. Capture sharpening is just the cake half or one-third baked. You need to do the Output Sharpening, and perhaps some Creative Sharpening targeted at any specfific issues.


Well, my question is about capture sharpening. I thought the major reason capture sharpening is generally done with a pretty light hand is that further adjustments affecting contrast could have an impact on the halos produced by USM sharpening. I guess I first need to know if this is true, and second--does this still apply if Focus Magic is not based on USM? Does that mean it does not depend on the illusion/perception of sharpening that is based on contrast?
Logged
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 6767


WWW
« Reply #127 on: October 09, 2007, 05:45:21 PM »
ReplyReply

Quote
Well, my question is about capture sharpening. I thought the major reason capture sharpening is generally done with a pretty light hand is that further adjustments affecting contrast could have an impact on the halos produced by USM sharpening. I guess I first need to know if this is true, and second--does this still apply if Focus Magic is not based on USM? Does that mean it does not depend on the illusion/perception of sharpening that is based on contrast?
[a href=\"index.php?act=findpost&pid=144925\"][{POST_SNAPBACK}][/a]

Hi Gloria,

I think Jeff would be best at answering the question about why capture sharpening is "light-handed". He'll correct me if I'm mistaken, but as I recall it, it's main objective is to correct primarily the mid-tones for the softening effect of the Camera's AA filter, and as a basis for building the remainder of the sharpening process. So it doesn't need to do any more than it does.  Turning to FM, According to its literature, FM is not based on USM; nonetheless, if one adds USM sharpening to it, based on what I've seen from FM alone, I believe one would need to be quite light-handed with both FM and the add-ons to avoid creating a brittle image.

Mark
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
Brian Gilkes
Sr. Member
****
Offline Offline

Posts: 431


WWW
« Reply #128 on: October 09, 2007, 06:23:19 PM »
ReplyReply

Not withdrawing  my previous comment, I think the issue of sharpening requires more research on at least 4 fronts
a/ chip designers need to reduce need for demosaicing e.g less space between pixel buckets and more precise blurring filters. This would improve absolute , if nor perceived resolution.
b/ RAW processors can concentrate less on detail removing smoothing which becomes less valuable as noise reduction improves.
c/ Printers need smaller drops (say 1.5 picolitres on large format printers)  and place the drops more precisely. There seems to be moves in this direction.
d/ Sharpening software needs to take account of deconvolution algorithms, which have definite advantages over local contrast enhancement. PKS is generally acclaimed as a great product. It seems to have the contrast bit nailed. The next step would be to look at incorporation of wave front reconstruction.For those coming our of caves, sorry darkrooms, deconvolution brings into focus out of focus images. Yes,Virginia,  that is possible  I have had considerable success with  Richardson-Lucy iterations and and am looking at Landweber, Myrheim & Rue, Gull and Daniel, and Weiner filtration. Successive iterations  caan provide astonishing results . For instance a completely out of focus image can be restored. The best known example is the restoration of Hubble telescope images. With ever swifter computers this technology bears investigation. What do you reckon Jeff?
Cheers,
Brian,
www.pharoseditions.com.au
PS I have an idea how to do it (the ultimate sharpening workflow), but  can't do code.
Logged
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #129 on: October 10, 2007, 01:33:59 AM »
ReplyReply

Quote
Jonathan, I use a 1Ds also and take many pictures like this, so I can say something about it - hopefully not too dumb. Firstly, I know that FM is using a deconvolution approach which is fundamentally different from what PKS, NIK and others do. But focusing on the results (pun intended), I'm viewing your comparison on a LaCie 321 display at 200% magnification. I find the edge detail on the roof of the building to the right of Gergen better with PKS. The building facades look a bit crisper and cleaner in the FM version, where the word "Merkur" is slightly more readable.

The roof texture is the only area of the image that PKS seems to do a better job than FM; in every other area FM does a better job IMO clarifying edges without introducing halos or excessively enhancing noise. If you look at the foliage in the bottom right corner and the building facades, FM definitely does a better job. My take is that deconvolution is simply a better approach than USM for undoing the blurring effect of an AA filter, which is the primary purpose of capture sharpening. Of course, things are different when you start talking about creative sharpening, which is generally local contrast enhancement on various spatial frequency ranges. Masked USM works great for that, so I would expect PKS to do much better comparatively than it has in this test.

Quote
try it again with Super Fine Edge sharpen and see what happens

I have, and the edge enhancement was even less effective than the medium setting. I tried the wide edge sharpen setting as well, and got some really ugly salt & pepper artifacts in the sidewalk. Overall the medium setting was the best of the bunch. You are welcome to take my posted image and try your own combinations of settings to see if you can improve on my results. I'm open to the possibility that I may not be using PKS to its full potential.
Logged

Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #130 on: October 10, 2007, 06:56:31 AM »
ReplyReply

Quote
For instance a completely out of focus image can be restored. The best known example is the restoration of Hubble telescope images. With ever swifter computers this technology bears investigation.

Focus Magic is an example of the technology, but such things have definite limits. You can take any B&W image and blur it until it is a featureless mass of pixels at level 128 or so. The software cannot look at the image and distinguish which image got blurred to the final featureless mass. You can enhance the detail that is there, but ultimately you cannot calculate detail that is not there.
Logged

Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 6767


WWW
« Reply #131 on: October 10, 2007, 07:21:36 AM »
ReplyReply

Quote
The roof texture is the only area of the image that PKS seems to do a better job than FM; in every other area FM does a better job IMO clarifying edges without introducing halos or excessively enhancing noise. If you look at the foliage in the bottom right corner and the building facades, FM definitely does a better job. My take is that deconvolution is simply a better approach than USM for undoing the blurring effect of an AA filter, which is the primary purpose of capture sharpening. Of course, things are different when you start talking about creative sharpening, which is generally local contrast enhancement on various spatial frequency ranges. Masked USM works great for that, so I would expect PKS to do much better comparatively than it has in this test.
I have, and the edge enhancement was even less effective than the medium setting. I tried the wide edge sharpen setting as well, and got some really ugly salt & pepper artifacts in the sidewalk. Overall the medium setting was the best of the bunch. You are welcome to take my posted image and try your own combinations of settings to see if you can improve on my results. I'm open to the possibility that I may not be using PKS to its full potential.
[a href=\"index.php?act=findpost&pid=145031\"][{POST_SNAPBACK}][/a]

Jonathan, I agree (and agreed above) that the building facades came out cleaner in Focus Magic. I also agree that the shrubbery in the FM result is sharper. However, there appears to be more preserved detail and texture in the roadway with the PKS result. That raises what may be the key issue in this comparison: the distinction between real image texture and noise.

It seems to me as if there is some cleaning-up associated with the FM workflow that hasn't happened - yet - with the PKS workflow, so could it be that the PKS workflow is sharpening some noise while the FM workflow is suppressing the noise and sharpening the edges?

I ask because the building facades themselves don't contain that much frequency which needs sharpening, and the edges they do have are rather wide and smooth. This does not pertain to the roofing, where there is higher frequency and sharper edges. Hence a workflow that embeds some noise reduction would make the facades look cleaner but not do quite as well on the high frequency detail. Does your FM workflow in this image include your actions after FM, or only FM?

I wouldn't jump to a conclusion that you are not using PKS properly - you may well be up to the point that you have implemented it, but one does need to complete the whole process to really see what it does.

Actually, this is becoming interesting and I wouldn't mind poking around your image to see what I can do with it. I'll send you a PM re transmitting the raw file.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
bjanes
Sr. Member
****
Online Online

Posts: 2714



« Reply #132 on: October 10, 2007, 11:10:17 AM »
ReplyReply

Quote
The next step would be to look at incorporation of wave front reconstruction.For those coming our of caves, sorry darkrooms, deconvolution brings into focus out of focus images. Yes,Virginia,  that is possible  I have had considerable success with  Richardson-Lucy iterations and and am looking at Landweber, Myrheim & Rue, Gull and Daniel, and Weiner filtration. Successive iterations  caan provide astonishing results . For instance a completely out of focus image can be restored. The best known example is the restoration of Hubble telescope images. With ever swifter computers this technology bears investigation. What do you reckon Jeff?
[{POST_SNAPBACK}][/a]

Brian,

Very interesting.

The Lucy-Richardson deconvolution of the Hubble telescope was relatively successful because the imaging defect was spherical aberration and it was possible to create a [a href=\"http://en.wikipedia.org/wiki/Deconvolution#Optics_and_other_imaging]PSF[/url] describing the effects of the aberration so that they could be removed by the deconvolution process.

As described in the link, the PSF for many images is not that easily derived. In your work, what implementations of the Richardson-Lucy or Weiner algorithm do you use, and how do you determine the PSF?

Bill
Logged
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #133 on: October 10, 2007, 11:42:08 AM »
ReplyReply

Quote
That raises what may be the key issue in this comparison: the distinction between real image texture and noise.

It seems to me as if there is some cleaning-up associated with the FM workflow that hasn't happened - yet - with the PKS workflow, so could it be that the PKS workflow is sharpening some noise while the FM workflow is suppressing the noise and sharpening the edges?

Based on my memory of the scene, I'd be inclined to say that in the areas where PKS appears to be finding more detail (roadway and rooftops), the detail isn't really there; it's definitely noise in the sidewalk/roadway and the front of the building, and the texture of the slate tile rooftop is definitely exaggerated compared to real life. The standard-issue German slate roof tiles are only about .5cm thick, and are the same color on the edges as the side, and under the flat lighting of the overcast day, the edges of the tiles shouldn't be nearly that blatant. It's an effect really more appropriate for creative or output sharpening, as it goes quite a bit past the mandate of simply undoing the softening effect of the AA filter during capture. Overall, Focus Magic's treatment is much more true to the subject.

Focus Magic has a noise reduction option in the dialog, which is set to Auto and disabled so that you can't change it. It doesn't do anything aggressive, but if you look at the FM vs original sidewalk area under the yellow striped awning, you'll see that there are a few noise specks in the sidewalk that are subtly de-emphasized after FM. It's very subtle and unobtrusive effect, but is there.

Quote
Does your FM workflow in this image include your actions after FM, or only FM?

The capture sharpening comparison I did involved FM vs PKS with the specified settings, no other tweaking. Since we've kind of gotten off the original topic here, I'm starting a PKS comparison thread so that discussion of Dan's action can continue without further interruption. I'll post a 16-bit TIFF of the Trier image I'm using for comparison in the new thread.
« Last Edit: October 10, 2007, 12:03:45 PM by Jonathan Wienke » Logged

I Simonius
Newbie
*
Offline Offline

Posts: 31


« Reply #134 on: October 22, 2008, 05:06:37 PM »
ReplyReply

Quote from: Schewe
As for your attempts at behavior modification bud, that's a lost cause...I'm way to old to care what others may think of me. It's just not relevant to who I am and what I do.


Your tone is quite an unpleasant surprise though Jeff
Logged
DarkPenguin
Guest
« Reply #135 on: October 22, 2008, 05:13:01 PM »
ReplyReply

Quote from: I Simonius
Your tone is quite an unpleasant surprise though Jeff

Decided to celebrate this thread's anniversary?
Logged
I Simonius
Newbie
*
Offline Offline

Posts: 31


« Reply #136 on: October 22, 2008, 05:35:41 PM »
ReplyReply

Quote from: Jonathan Wienke
My negative comments directed toward Dan and his action have several purposes:

4. Ultimately, to save others the time and hassle of wasting time fooling with sharpening methods that have problems that were solved years ago by much better techniques.

and for that I thank you  

- also to everyone else for the wealth sharpening info here:)
Logged
I Simonius
Newbie
*
Offline Offline

Posts: 31


« Reply #137 on: October 22, 2008, 05:38:21 PM »
ReplyReply

Quote from: DarkPenguin
Decided to celebrate this thread's anniversary?

Could do - by all means:)

Just came across it by chance - fascinating read for sharpening info, difficult not to be shocked by the personal stuff though, however justified the posters might have felt.
Logged
bjanes
Sr. Member
****
Online Online

Posts: 2714



« Reply #138 on: October 22, 2008, 09:06:19 PM »
ReplyReply

Quote from: DarkPenguin
Decided to celebrate this thread's anniversary?

Has anything changed in the past year? For users of Adobe Camera Raw, capture sharpening has been added or improved (I can't remember for sure when it was released) and ACR users might want to use its sharpening for comparison rather than that of PKSharpener. Jeff Schewe, one of the developers of PK, says that ACR's capture sharpening is superior to that of PK. Has anything changed in PSCS4's smart sharpen--the interface looks similar, but there could be internal improvements.

Deconvolution image restoration (such as used in Smart Sharpen and FocusMagic) theoretically should be superior to unsharp masking, since it can actually correct defocus, diffraction, and possibly the image degradation from the camera blur filter. However, it is difficult to derive the necessary point spread function necessary for the deconvolution, and it seems as if relatively few photographers use deconvolution outside of astro-photography. Anything new here?

Bill

Logged
jjj
Sr. Member
****
Offline Offline

Posts: 3078



WWW
« Reply #139 on: October 22, 2008, 09:33:42 PM »
ReplyReply

Quote from: bjanes
Has anything changed in the past year? For users of Adobe Camera Raw, capture sharpening has been added or improved (I can't remember for sure when it was released) and ACR users might want to use its sharpening for comparison rather than that of PKSharpener. Jeff Schewe, one of the developers of PK, says that ACR's capture sharpening is superior to that of PK.
Did he? Where? If so I'm impressed by his honesty, not that he's been known for not being very honest and forthright in his opinions!  
Logged

Tradition is the Backbone of the Spineless.   Futt Futt Futt Photography
Pages: « 1 ... 5 6 [7] 8 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad