Dan Margulis Sharpening Action

[digitaldog]

One big problem I'm having trying to access if Dan's technique is any good is figuring out what its for. As I said, is it capture sharpening (well based on running it, it can't be. More later). Is out output sharpening? Anyway, I downloaded the action from DP Review to run the sharpening technique. I took two very different images: one high key, one low key night shot at a pretty high ISO and ran the images using the defaults (lets not even go into the fudge factor instructions provided by Dan about "if this looks too much or too little).

I wanted to compare this to PhotoKit Sharpener but using what? After viewing the images at 100% in  Photoshop from Dan's technique, there's no way this is gentle capture sharpening. Its way, way too sharp for that. So how do I compare it to PKS? Well I used output sharpened for Ink Jet, 300ppi and glossy paper. Anyone who's familiar with this product will know that when you view the effect of this kind of output sharpening on screen, it looks prettly awful but prints fine. Its a wake up call to how silly it is to use a low rez device like a display to visually approach output sharpening. Below I have two examples, one with PKS, one with Dan's technique. Dan's makes the PKS output sharpening look like gentle capture sharpening. What's really really bad here is what he's doing to low key images. They are hosed! And yes, I printed them both out to a 3800 on Luster paper and yes indeed, the severe damage you see in Dan's technique shows up like a sore thumb.

Only two files gang but based on these tests (from Raw, no sharpening in CR, 16-bit ProPhoto RGB), I can only say this technique looks like garbage! Other than the fact the image IS sharpener, its about the worst technique I've tried on images for sharpening in a very long time.

As it appears I can only post one image per post, the first is the night shot, side by side screen dump at 100%. I think you can easily tell which is which. I only wish you could see the prints!

Next post will be my girl Zia in the snow (high key).

It appears that this technique is a bit more forgiving on lighter tones and totally hoses dark tones. But if you look at the high key stuff, it looks pretty darn ugly. But its hard to say anything definitive about this test since we have NO idea from Dan when or how to use this, other than run the action and tweak.

[digitaldog]

High key...

[Mark D Segal]

Both images are over-sharpened to the point of unpleasant halos and excessive white artifacting - this is so obvious on the display there is no question in my mind they would print similarly. I haven't seen your prints of course, but I'll stick my neck out and hazard a guess they probably look like those low-res, over-sharpened JPEGs that came out of the digicams from the 1990s - "eh" ? (I'm Canadian, gotta get that "eh" in there!)

[digitaldog]

Both images are over-sharpened to the point of unpleasant halos and excessive white artifacting - this is so obvious on the display there is no question in my mind they would print similarly.[a href=\"index.php?act=findpost&pid=143694\"][{POST_SNAPBACK}][/a]

Well the output sharpening from PKS, as I said, usually looks pretty awful. That's to be expected. I could upload the 'before' (no sharpening) but again, what readers here have to do is run the action and output the files on their devices. What really looks absolutely awful in Dan's techniques is what's happening in low frequency, almost smooth areas of dark tones. Those are NOT stars in the night shot <g>. In the fur of the darker regions of the dog, there's also some nasty artifacts.

But in the end, this is almost pointless. I'm assuming I ran the action correctly. There are two dialogs that pop (USM and Levels), and I'm supposed to do something here but I have no idea what or based on anything specified. Dan or a minion could say I did this incorrectly. Considering there's no more instructions than 'run the action', I'm at a loss. Then there's the fact I'm comparing this to PKS output to an Ink jet glossy paper 300dpi. Is that fair? Well since I have no idea what Dan's action is for, it may or may not be. If its based on halftone output to a press, and I use PKS for ink jet and print both side by side, I didn't run a fair test and gave PKS an unfair advantage.

Yes, Dan's action does make the image look incredibly awful on screen. But based on what I know about visual sharpening, I guess it could be optimal for something of which hasn't been defined. As such, there's really no fair way to test this other than using the "it looks sharper after then before" which is silly.

Visual sharpening is fraught with issues as many of us know. Our brain plays tricks on us. Over sharpen an image, then fade it back or move the sliders back, the image visually looks softer. I suspect this is one reason so many over sharpen. You have to train yourself to look for areas like harsh halo's or artifacts and not over do it. BTW, I found the same 'optical' effect when using noise reduction. I'm a fan of NoiseWare, and like PKS, I usually let it do its preset thing, rather than try to out smart it. When I move the various sliders to remove the noise, the effect visually is that image looks less sharp. Only when you output the before and after do you see what a great job it does. Certain edits just don't work visually on a display. Sharpening and noise reduction can fall into this camp.

Anyway, I do know that despite what an image looks like at 100% on any display, when I use PKS and output the files, they look as I wish. But I conduct this sharpening based on a well defined methodology and based on what the heck I want to do with the image. None of this applies to Dan's technique. From my perspective, its pretty useless.

Considering Dan's MO in the past, the totally vague instructions for use, with the advise that you should tweak based on visual feedback, this seems like something we might have given a few minutes of attention to way back in Photoshop 2 days. Unless Dan can provide some kind of instructions about what the heck this technique is for, I have to believe its yet another 32 step turd polishing routine that is half baked, not tested. But I'm biased. I fully admit that.

[Mark D Segal]

Andrew,

Did you make prints of the results you got from both Dan's proedure and PKS on those two files you posted? Or are you discussing on the basis of the display view? My take on Dan's procedure is that it is "sharpening" full stop. Like before the days of a "multi-pass sharpening workflow". If so, there's not much point trying to match it with either Capture or Output sharpen. The only valid comparison would be actual prints resulting from running on the one hand Capture + Output in PKS, and on the other hand Dan's procedure. Then compare the prints.

Mark

[digitaldog]

Yes, printed. But again, when comparing to PKS, I used a method based on the output I was using (an Epson 3800 on luster paper).

On the print, it doesn't of course look as butt ugly as we see here. But I do see what looks like white pixels (stars) in the Dan version in the sky where in the PKS, its clean and black as the original image shows. I also see noise greatly exacerbated in Dan's image in smooth areas of tone which PKS avoids. In print, there's nothing I see in Dan's version I like over the PKS version, just the opposite. Some might say "Dan's version of the print looks sharper' because it is. The there's far more noise but when you look at edge detail, its no better. And that's where we want to see the effect of sharpening.

Do we really believe that this is output sharpening that's appropriate for any output device, from any resolution file?

And IF I really want to stack the deck in PKS's favor for visual analysis, I could have used output sharpening for the web and posted the images side by side. That would really make Dan's version look worse.

This effect of Dan's looks like the over-sharpened drum scans we got in the early 90's. This might print well in a book or to a press. But since I don't know what its based on, all I can say is that to a good ink jet printer, its not at all useful.

[laughfta]

I went back to look at Dan's directions for doing this sharpening. They are:


Here's the procedure, which of course should be reduced to an Action to save having to do
it over and over.

1) Copy the RGB image.

2) With the copy, Convert to Profile>Custom CMYK.

3) Fill in: Heavy GCR, 70% black ink limit, 340% total ink. Dot gain is basically not relevant
as you can always lighten or darken the mask after applying it, but I just use the default
20%.

4) Click OK twice to generate the false separation.

5) Command-4 to expose the black channel, and Mode: Grayscale to discard the CMY
channels.

6) Invert the channel with Command-I, yielding a negative image.

7) Auto Levels.

Gaussian Blur, radius 2.0 pixels to eliminate noise and make for a softer sharpen.

9) Return to the RGB image and create a duplicate layer. Sharpen conventionally with a
very heavy hand--500%, 1.2 pixel Radius, 3 Threshold might be a good starting point for
most images.

10) Add a layer mask. To it, load the artificial black channel that was made in steps 1-8.
This should confine the sharpening to the desired areas.

11) If you feel the image is not sharp enough, apply a curve to the mask to lighten its
midpoint. If you find the image to be too sharp, darken the mask in the same way.



So I did this, exactly as written (no luminosity mode change), and I ended up with a file that looks...fine. Certainly workable. Was too sharp, but I darkened the mask and  it seemed much better. I'm not sure how I feel about the print, but I can say that this  technique does not belong in some crazy sideshow; it could be viable under some circumstances.

I have no idea how large a file to attach, but here goes.

[laughfta]

I don't know how to prepare a file for viewing here, obviously. Poor Dan  

[digitaldog]

11) "If you feel the image is not sharp enough, apply a curve to the mask to lighten its
midpoint. If you find the image to be too sharp, darken the mask in the same way".

So I did this, exactly as written (no luminosity mode change), and I ended up with a file that looks...fine. Certainly workable. Was too sharp, but I darkened the mask and  it seemed much better. I'm not sure how I feel about the print, but I can say that this  technique does not belong in some crazy sideshow; it could be viable under some circumstances.

I have no idea how large a file to attach, but here goes.
[a href=\"index.php?act=findpost&pid=143718\"][{POST_SNAPBACK}][/a]

That's kind of the problem. IF you feel its too sharp. Based on what? If the idea is to print something, find its too sharp and tweak, well that's not at all useful, go back to good old and simple USM and fumble around. Might as well build a sharpening Variation's, output the squares and pick (hey, neat idea). But again, with respect to this set of commands, where's the beef?

Not sure how you feel about the print? Can you compare this to another sharpening routine, even a demo of PKS if you don't have it? It runs fully functional for 7 days.

It would be interesting to output the same image to different types of printers and papers. Will they all appear the same? Doubt it. Which is right?

No one is arguing that this produces a sharp (sharper) print. But then any USM setting would do that.

[laughfta]

That's kind of the problem. IF you feel its too sharp. Based on what? If the idea is to print something, find its too sharp and tweak, well that's not at all useful, go back to good old and simple USM and fumble around. Might as well build a sharpening Variation's, output the squares and pick (hey, neat idea). But again, with respect to this set of commands, where's the beef?

You know, I ended up feeling like the print was much less delicate than I'm used to seeing. Hard to explain, but though there was no obvious artifacts, there was a certain clarity missing; it feels heavy.

 

Can you compare this to another sharpening routine, even a demo of PKS if you don't have it? It runs fully functional for 7 days.

I think I'll do that tomorrow. I will say, this method was wasted on me--I don't have the know-how to make it work the way he intended. I only tried it because I couldn't believe it was completely insane.  

It would be interesting to output the same image to different types of printers and papers. Will they all appear the same? Doubt it. Which is right?

Tell you what, I'll print it after the Print Academy Workshop I'm attending in NY!

[Jonathan Wienke]

So Andrew, would you say that my initial assessment (any sharpening method that increases in intensity as you go down the tonal scale is flawed) is accurate?

[Schewe]

So Andrew, would you say that my initial assessment (any sharpening method that increases in intensity as you go down the tonal scale is flawed) is accurate?
[a href=\"index.php?act=findpost&pid=143728\"][{POST_SNAPBACK}][/a]

I do...one of the primary goals for PhotoKit Sharpener is to roll off the extreme highlights and the shadows and concentrate the sharpening in the midtones...to roll off the midtones and higher is, well, let's call it misguided...particularly for digital capture.

[Eric Myrvaagnes]

I used to fuss around with Unsharp Mask when I was trying to learn digital, and sometimes my results weren't too terrible. But once I tried PK Sharpener, I have never found any reason to go back, or to try any other method (or recipe or arcane ritual). Unless someone can show me an image that prints better after Dan's procedure than after capture+output from PKS, I see no reason to waste the time running experiments.

I, too, feel that Dan should provide the evidence (concrete examples, with details of how his "method" was applied) if there are indeed any situations in which his procedure is better than others (such as PKS defaults).

As you can see, I'm already getting biased against Dan's procedure, mainly because Jeff, Andrew, and Mark all provide explanations and the evidence that I haven't seen from Dan. I'll listen to people who make sense and explain what they are talking about without a lot of vague hand waving.

[Schewe]

hand waving.
[a href=\"index.php?act=findpost&pid=143737\"][{POST_SNAPBACK}][/a]

Hey, I'm pretty good at hand waving...move along, there's nothing of interest to you here.

:::waving hand in front of the stormtrooper's face:::

[sniper]

I went back to look at Dan's directions for doing this sharpening. They are:
Here's the procedure, which of course should be reduced to an Action to save having to do
it over and over.

1) Copy the RGB image.

2) With the copy, Convert to Profile>Custom CMYK.

3) Fill in: Heavy GCR, 70% black ink limit, 340% total ink. Dot gain is basically not relevant
as you can always lighten or darken the mask after applying it, but I just use the default
20%.

4) Click OK twice to generate the false separation.

5) Command-4 to expose the black channel, and Mode: Grayscale to discard the CMY
channels.

6) Invert the channel with Command-I, yielding a negative image.

7) Auto Levels.

Gaussian Blur, radius 2.0 pixels to eliminate noise and make for a softer sharpen.

9) Return to the RGB image and create a duplicate layer. Sharpen conventionally with a
very heavy hand--500%, 1.2 pixel Radius, 3 Threshold might be a good starting point for
most images.

10) Add a layer mask. To it, load the artificial black channel that was made in steps 1-8.
This should confine the sharpening to the desired areas.

11) If you feel the image is not sharp enough, apply a curve to the mask to lighten its
midpoint. If you find the image to be too sharp, darken the mask in the same way.
So I did this, exactly as written (no luminosity mode change), and I ended up with a file that looks...fine. Certainly workable. Was too sharp, but I darkened the mask and  it seemed much better. I'm not sure how I feel about the print, but I can say that this  technique does not belong in some crazy sideshow; it could be viable under some circumstances.

I have no idea how large a file to attach, but here goes.
[a href=\"index.php?act=findpost&pid=143718\"][{POST_SNAPBACK}][/a]

Has anyone compared the action to the above quote? I'm wondering if theres any difference between the two?   Wayne

[Mark D Segal]

You know, I ended up feeling like the print was much less delicate than I'm used to seeing. Hard to explain, but though there was no obvious artifacts, there was a certain clarity missing; it feels heavy.

 

I think I'll do that tomorrow. I will say, this method was wasted on me--I don't have the know-how to make it work the way he intended. I only tried it because I couldn't believe it was completely insane.  
Tell you what, I'll print it after the Print Academy Workshop I'm attending in NY!
[a href=\"index.php?act=findpost&pid=143727\"][{POST_SNAPBACK}][/a]

Gloria, if the method is "wasted on you" don't assume that the problem is you - it is more likely the method. I won't know this first-hand till I try it myself but that won't be till later this month, if at all, based on what is now emerging. But your results are consistent with those of several others.

In addition to which, when we now have plug-ins such as PK Sharpener with such user-friendly and intuitive UI, plus an extensive well-written manual for ease of reference, that sets a standard for new ideas - indeed I share EricM's philosophy on this point.  

I also share your philosophy of sharpening - not too much of a good thing. Make the image breathe with detail, without being harsh and brittle. As well, when doing volume work, the sharpening shouldn't be a hassle to adjusst for each image - yes, each image may need some custom-tweaking, but the algorithm should be designed to make that quite easy. All of these "expectations" (of mine) are amply fulfilled with PK Sharpener, which is why I use it.

Looking at the image you posted yesterday, if you care to do a comparison using that image, I would suggest that your Capture Sharpen be set to either Medium Edge or Narrow Edge,  and the Output Sharpen to suit the media and resolution of your print. Please merge the sharpening layers into the background after each stage of sharpening (you don't need to save them flattened - this is just for printing purposes) because those sharpening layers add hugely to file size and you don't want to send so much data to the printer. My normal workflow is to flatten capture sharpening into the background, because I have enough confidence in what I'm doing with capture sharpening (and it is gentle) that I don't expect any need to review it. Thousands of images later confirms this. I delete Output Sharpening after printing, because if ever the file were repurposed (resolution or medium) Output Sharpen would need to be re-done differently anyhow.

May I suggest, if you have the time, to select an image that has a bit more frequency - i.e. something with an area or two in the lights and darks which contain finer texture and detail you want emphasized but not over-emphasized.

[picnic]

I've followed this thread for days now--I honestly didn't have any desire to try Dan's method, but I find I usually learn something from these long technical threads.  

The only thing I have to add here--and it really is whining LOL--is that I am now using LR most of the time and I have to do a round trip to PS to use PKS (or else just print from PS)--and ever since I started using it I haven't used any other form of sharpening (except to try the sharpening in LR)--why try to do something on your own that won't be nearly as good as the relatively simple PKS which also offers so many options if you choose (creative).  I have great hopes that a form of PKS can be added somehow to LR in the future.

Diane

[Mark D Segal]

I've followed this thread for days now--I honestly didn't have any desire to try Dan's method, but I find I usually learn something from these long technical threads. 

The only thing I have to add here--and it really is whining LOL--is that I am now using LR most of the time and I have to do a round trip to PS to use PKS (or else just print from PS)--and ever since I started using it I haven't used any other form of sharpening (except to try the sharpening in LR)--why try to do something on your own that won't be nearly as good as the relatively simple PKS which also offers so many options if you choose (creative).  I have great hopes that a form of PKS can be added somehow to LR in the future.

Diane
[a href=\"index.php?act=findpost&pid=143794\"][{POST_SNAPBACK}][/a]

Yes indeed Diane - that and soft-proofing.

[digitaldog]

Has anyone compared the action to the above quote? I'm wondering if theres any difference between the two?   Wayne
[a href=\"index.php?act=findpost&pid=143755\"][{POST_SNAPBACK}][/a]

First differences is convert to 16-bit in the first step which seems anti-Dan.

There's a step in the action that makes sense to me that I don't see in Dan's original post (nor yours) where after applying the G-Blur, and duplicating the bkgnd layer, layer styles are called forth and the Blend If settings are applied:5/10 242/250. Without this step, the results should be even more harsh.

The action has a lower USM setting! Dan recommends 500%, 1.2, Radius 3. The action had 500/0.8/3! So the sharpening I was seeing was less than specified (unless Dan instructed this guy differently). I have to do my tests over again but I suspect the results will be even worse since I'll undo the Blend If and up the USM.

[digitaldog]

Guess what, the troll Iiah just posted a NEW action to the DR review list to 'fix' an error message. The URL is below if you want to grab it. I posted this just now, as he's still not following Dan's recipe (it looks like he's trying to make it less ugly using his action, hence the Blend If and lower USM settings). In his new action, the Blend If is still there, the USM is back to Dan's spec (higher, I suspect not for the better).

liah Borg wrote:
> Here is a version that won't cause that message
>
> http://www.pochtar.com/DanMargulis_RGB_Sharpen_K_mask_v2.zip

UNLESS Dan's told you otherwise, you're NOT following this recipe as specified. Your first download changed the USM settings. You set them for 500/0.8/3. WHY? He species 500/1.2/3. The new action now follow Dan's now (higher values, yuck). But what's with the Blend If settings in the duplicate layer step? He doesn't specify this either. All the testing many of us have done (providing pretty awful results on low key images) are not valid using your action. At least until you tell us that Dan instructed you to alter his settings. He also doesn't specify conversion to 16-bit as your does. WHY?

Talk about bad science again.

So I'll start again, but based on the exact settings by Dan (no blend if, higher radius), I can't believe I'll see anything but more specks in my black sky.