State of the art sharpening workflow?

rutt · April 8, 2008

Since I wrote my sharpening tutorials a few years ago, people seem to have developed some new ideas about when and how to sharpen and how to fit this into their workflows. In particular, PF says he now uses ACR for "capture sharpening", obviously early in the workflow.

I'm afraid I've fallen way behind. I really don't know anything about this, and I'm sure I'm not the only one. What about a nice easy to understand explanation of the rational, theory, and practice here? What has changed since I thought I understood this topic?

Thanks.

pathfinder · April 8, 2008

John,

This is a interesting subject with lots of ramifications and some differences of opinions as I am sure you are aware.

For "capture sharpening" - to recapture the sharpness lost in digitization ( with typical anti-aliasing filters built into most digital cameras - Sigma's Foveon being an exception ) I follow what I learned from Jeff Schew and Michael Reichman in their video from "From Camera to Print" and Schew's and Fraser's book "Real World Camera RAW for Adobe PSCS 3". These are both excellent resources and bare careful attention. These same techniques I also saw demonstrated by Marc Muench at the Glacier and Utah workshops last year.

The ability to sharpen with a mask to limit sharpening artifacts, inside ARC, is a very powerful tool.

I will be attending Marc's print workshop next week - so I would like to delay my full post until after that - I may find Marc has some new tricks to teach in the mean time

Welcome back - I always learn new things from your efforts.

Travis · April 8, 2008

This will be a very interesting thread to follow. I'm finally starting to gain an understanding of what capture sharpening is. I use Lightroom instead of ACR (same controls and I hope the same algorithm). My area of question is how capture sharpening and masking to apply without going to far. The goal is not to do display or print sharpening. I'm looking forward to reading more on this topic....

rutt · April 8, 2008

Thanks, Jim. I'm sure there are differences of opinion here, but I was hoping to learn something without having to dig though too much information. When I wrote those tutorials, I thought I understood the subject pretty well. Now I'd like to find out if I can improve on my technique from that time. If I do learn something, perhaps I can update the tutorials to reflect current best practices.

One thing I have learned since those tutorials is to sharpen though various masks. For example, I sharpen the ballet photos through a steepened K channel mask in order to avoid sharpening noise. Dan Margulis has been experimenting with sharpening through an inverted K channel mask to avoid sharpening bright colors. I've been meaning to post this idea; given the state of memory these days, perhaps I already have.

I tried using ACR's sharpening and didn't really manage to understand how it works. What is the mask? How is it derived? My tutorials embodied Dan's USM workflow:

Maximize amount, minimize threshold so you can see what you are doing.
Choose a radius that doesn't obscure detail.
Choose a threshold which doesn't sharpen noise or other unwanted fine details.
Choose an amount which provides the effect without obvious halos.

I'd love to learn something as simple as this for ACR sharpening.

colourbox · April 8, 2008

rutt wrote:

I tried using ACR's sharpening and didn't really manage to understand how it works. What is the mask? How is it derived? My tutorials embodied Dan's USM workflow:

Have you already checked out the ACR sharpening explanations here and do they provide enough info to update the workflow? Or were you looking for a deeper explanation?

arodney · April 8, 2008

Travis wrote:

I use Lightroom instead of ACR (same controls and I hope the same algorithm).

Yes.

My area of question is how capture sharpening and masking to apply without going to far. The goal is not to do display or print sharpening. I'm looking forward to reading more on this topic....

The controls in Develop are such that you can't do too much damage unless you go totally overboard. It is visually based which has issues. The masking is really no different from what we'd manually build in Photoshop, only faster, easier and all being done non destructively and in linear encoded color space (better masking of highlights possible here). Like all sharpening, you're better under than over doing it. Now with output sharpening in 2.0 (based on the capture sharpening and what kind of resolution and size you'll print to, along with paper type and three strengths), beginning to end sharpening is pretty darn easy (which is what it should be). Unless you just like to waste time and dink around, there's no reason to worry about this stuff in Photoshop. Exception, creative sharpening.

The bases for LR.'s sharpening is from Bruce Fraser: http://www.creativepro.com/story/feature/20357.html

sesshin · April 8, 2008

For a while now I've been using Bruce Fraser's sharpening method laid out in "Real World Image Sharpening", but have found the controls in ACR to be a great alternative. I love the masking ability. I used to spend so much time creating custom masks manually in PS, its a joy to have them generated automatically.

The only drawback, from what I can see, is that you can use curves on a manually created mask in photoshop, effectively steepening the transitions between sharpened and unsharpened areas, and you can't do this in ACR. Its still great for what it is though.

BinaryFx · April 9, 2008

rutt wrote:

I'm afraid I've fallen way behind. I really don't know anything about this, and I'm sure I'm not the only one. What about a nice easy to understand explanation of the rational, theory, and practice here? What has changed since I thought I understood this topic?

Thanks.

John, this is a good question, with many possible answers and opinions!

This reply is from a prepress perspective, which may differ a bit from photographic workflow. It may also be a good idea to look to the past and the present to see how things are developing, as a possible guide to the future.

Back in the day, the USM was applied for press output in the drum scanning stage (originally analogue then digitally as time went on). Results were often dependent on the operator as to how much sharpening was applied, although it was generally targeted for halftone repro at final print size (degree of USM for light and dark contours, contour width, threshold etc). Scaling of the image in layout was often no more than say 20% up or down if the image was not being used at the scan size originally indicated in the layout [the old FPO flag and all that, for the old timers out there!].

When flatbed scanners mostly took over the bread and butter work of scanning, some still applied output USM at the scan stage as before - while some others did not apply USM in scanning and did it later in Photoshop.

This is when many of the USM tricks and techniques started, then with layers in v3 with layer masks and blend if sliders were added.

It became common for folk to apply a small acquisition/capture USM, to account for the scanning process softening the image slightly. Some creative sharpening of key elements may have also been applied. Minor sharpening was also used after resampling down. Then the image would get a final round of output targeted sharpening. This may or may not have been done with care or skill (depending on the person) and progressive rounds of sharpening may have amplified problems introduced in previous sharpening rounds.

Bruce Fraser was perhaps the first to make this multi-stage sharpening approach into a popular published "workflow" - where he used various tricks to try to keep the early stages of sharpening from producing negative results when the final output sharpening was applied. The popular PhotoKit Sharpener automation plug uses these principles (edge masks, variable light dark contour/halo control, blend if sliders, luminosity blending etc).

You asked for a brief explanation...hah!

Anyway, an early piece on a multi-stage sharpening process can be found here from the late Mr Fraser:

http://www.creativepro.com/story/feature/12189.html

Over time, this became a 3 step process and Bruce refined things further and wrote about it in books and implemented it in the commercial PKS automate plug. Fast forwarding to the present, Adobe are adding these common sharpening tricks to ACR/ALR.

Usually some previous acquisition and or creative sharpening is helpful when resizing the master image dupe down to final output size for output sharpening. Sometimes a sharper file may create problems when resizing some problematic fine detail down. Making the image softer, not sharper, actually resolves this issue!

If one is printing at same size, then some content may be too crisp with output sharpening applied over capture/creative sharpening. In these cases, one may have to trial reducing opacity of or removing the sharpening layers before appliction of the output sharpening to the duped output image.

Deconvolution based focus blur or motion blur restoration is a newer area than classic USM that is gaining some ground in graphic arts (Hubble telescope is a good example from astronomy). In CS2, Adobe added Smart Sharpen, which disappointed me, as I was expecting more based on what I have seen possible from other sources. I hope that Adobe can do more with this, as it appears to be a good way to handle input/acquire/focus issues, which are often different to USM (although USM can be used to some effect).

There are many ways to apply "sharpening" to an image, with the high pass method being a popular alternative to USM (as many ways as to blur or soften an image). It is possible to sharpen with the different blur filters, median filtering, by resampling etc.

Many of the painstaking tricks employed in ACR/ALR or Photoshop or elsewhere matter for little when it comes to ink hitting paper in say an inkjet printing situation on semi-gloss stock...the proof is in the final print, not the monitor. On the final print at intended viewing distance, simple USM often looks very similar to the "power user/tool" sharpen that pros slave over with all the tricks (when output at same size and not enlarged).

Hope this helps,

Stephen Marsh
http://members.ozemail.com.au/~binaryfx/

arodney · April 9, 2008

sesshin wrote:

For a while now I've been using Bruce Fraser's sharpening method laid out in "Real World Image Sharpening", but have found the controls in ACR to be a great alternative. I love the masking ability. I used to spend so much time creating custom masks manually in PS, its a joy to have them generated automatically.

Yup, but some don't feel that's macho (or they have way too much time on their hands). They would rather introduce a 37 step process of hoops to jump through, almost always based on some visual feedback (silly, considering the huge disconnect between the resolution of a display and a modern printer). The only thing lacking in LR 2.0 (and it might come) is selective sharpening and blurring to provide creative sharpening. Note that using a negative value for Clarity in the new brush in 2.0 will provide some slight blurring effect. Otherwise, in 2.0, the creative and output sharpening are providing everything tasty I see in PKS done slower and by making the files larger in Photoshop.

The masking as you point out in LR is really great, easy to use, flexible and breaks all the old Photoshop rules (LR is as the chief designer likes to call it, the anti-Photoshop). The Photoshop team can (and IS) learning from the LR team (we'll all have a nice conversation about this when CS4 ships <g>). There are so many ways to accomplish the same goals easier, faster and non destructively with more modern tools like Lightroom and ACR.

The output sharpening in 2.0 beta is still undergoing tweaking so having feedback will be useful to the team (and Jeff Schewe who's helping them). Also, it looks like a contone setting (for Lambda, Lightjet) might be useful. The sharpening is also applied in the Web gallery and I think its going to get hooked up for Slideshow too.

So short of CMYK (leave that to Photoshop), Bruce's work with respect to sharpening has found most of its way into Lightroom. It will be a cold day in hell when we see CMYK or other color models in LR/ACR. That's the KISS approach to keeping that app focused and streamlined (thankfully).

rutt · April 9, 2008

Andrew, I plead guilty. I do sometimes sweat bullets to get what I want from a picture. Sometimes that's because I think it's worth it because the picture is something special and I want it to be the best I can make it. Sometimes, it's because I'm just enjoying myself. And most often it's because I'm trying to learn something new (as now.) The easiest thing for any of us to do is just to keep doing the same thing we already know how to do. I have some "37 step process" complete with hoops and all which I feel comfortable with. I've done it so often that it's pretty darn quick for me now. In fact, it's much faster than learning to do it all in ACR or LR. Given the way I learn, that will take me quite a bit of time to feel comfortable with.

But I like learning new things, especially when I can put them to use and get improved results (including less work to get the same or better quality.) So I opened up this thread to see what I could learn. It will make it easier for me to do that if you don't hurt my feelings by disparaging what I have already learned.

Thanks.

arodney · April 9, 2008

rutt wrote:

The easiest thing for any of us to do is just to keep doing the same thing we already know how to do.

Actually the newest way to do this (not just doing same thing) was just introduced a week ago. Fortunately, its real easy too.

rutt · April 9, 2008

arodney wrote:

Actually the newest way to do this (not just doing same thing) was just introduced a week ago. Fortunately, its real easy too.

Pointer, please. It's never going to be easy for me if I don't know where to find it.

And believe me. Old dogs (like me) don't learn new things easily. Really, for me the easiest thing is the thing I know and understand already.

One man's meat is another man's poison. -- Aesop

arodney · April 9, 2008

rutt wrote:

Pointer, please. It's never going to be easy for me if I don't know where to find it.

You can start here:

http://lightroom-news.com/2008/04/02/lightroom-2-beta-available/

pathfinder · April 9, 2008

rutt wrote:

Thanks, Jim. I'm sure there are differences of opinion here, but I was hoping to learn something without having to dig though too much information. When I wrote those tutorials, I thought I understood the subject pretty well. Now I'd like to find out if I can improve on my technique from that time. If I do learn something, perhaps I can update the tutorials to reflect current best practices.

One thing I have learned since those tutorials is to sharpen though various masks. For example, I sharpen the ballet photos through a steepened K channel mask in order to avoid sharpening noise. Dan Margulis has been experimenting with sharpening through an inverted K channel mask to avoid sharpening bright colors. I've been meaning to post this idea; given the state of memory these days, perhaps I already have.

I tried using ACR's sharpening and didn't really manage to understand how it works. What is the mask? How is it derived? My tutorials embodied Dan's USM workflow:
Maximize amount, minimize threshold so you can see what you are doing.

Choose a radius that doesn't obscure detail.

Choose a threshold which doesn't sharpen noise or other unwanted fine details.

Choose an amount which provides the effect without obvious halos.

I'd love to learn something as simple as this for ACR sharpening.

Capture Sharpening is what ACR does - restores the sharpness lost due to digitization, and the blurring from the anti-aliasing screen in front of your sensor. (The anti-aliasing screen is there to minimize moire' by blurring the image on the sensor slightly)

John, you sharpen your images with masks to precisely control where you sharpen, and to avoid sharpening noise in the shadows. That is exactly why I like sharpening in ACR 4.3, but without the aggravation of finding, building and moving masks around. For me, initial sharpening in ACR ( or LR) is much faster. Whether it is truly 'better' I do not know, but I do know that my images are acceptibly sharp to my eye, even at very large sizes. I see no sharpening halos either. (I will state that I have very little interest in trying to make acceptable images from blurred shots that should be deleted and not processed. My interest is to make what I feel are very good initial images even better)

When sharpening in ACR 4.3 et al, it is critical to have your image at 100% or larger - so that one can see the B&W luminosity image that is being processed. This is very important ( And I suspect your are, John, but not all of my readers will be, I suspect)

The following answers are derived directly from Jeff Schewe's previously mentioned article

Amount is a 'volume' control that directly controls the strength of sharpening from 0 ( no effect ) to 150. I rarely use more than 115 or so and usually less than 95. I almost always use 50 or 60 or higher.

Radius is the number of pixels on either side of the selected 'edge' that sharpening is applied to. I rarely change that much from the default of 1.0. It depends to a degree on the image size and whether the image is a high frequency image or a low frequency image. I think that increasing this significantly increases the risk of halos, but would like to hear A Rodney's or others opinions about this.

Detail is similar to Threshold for USM, but "not exactly" according to the author. At a value of 0, Detail will limit halo dampening on the sharpening, or will almost remove the haloes completely. Running Detail all the way to the right ( 100) creates an effect very much like USM. I still find Detail the hardest to understand, but typically use a value between 18 and 30. I will be interested to hear other's opinions on Detail. The Detail control allows applying more sharpening without the light and dark halos typically seen with USM, and that rutt and others are using masks to help control or eliminate in Photoshop sharpening.

Masking: Masking creates an 'on the fly' edge mask that looks like a high contrast B&W negative - black conceals, and white reveals. The value needed is a function of the fine detail in the image, and the other parameters chosen. I like to use a fairly strong value at times to limit sharpening only to the major 'edges' of the image. A value of 0 for Masking results in no masking and the sharpening is applied globally across the image. I frequently use values from 10-15 up to 50 or 60 or even more. I rarely, if ever, use zero.

The article by Jeffe Schewe does not describe exactly how ACR fabricates the mask. I suspect it is proprietary with Adobe.

Reading between the lines in B Fraser's "Real World Sharpening with Adobe Photoshop CS2" may give some hints. On page 125 -126 he discusses using a red or green channel as a mask. He continues, that it is a small step to using Channel Mixer to create an even better mask, just like creating a better B&W image to blend with a color image for better control of contrast and color via the Apply Image command. The mask also needs to be inverted after Channel Mixer, and then blurred a bit with Gaussian Blur, and maybe receive a final curve to steepen the contrast of the mask as well. I have seen no discussion of what the mask truly represents, but these pages in B Fraser's "Image Sharpening in Photoshop CS2" sure seem to me a very likely starting place.

When using the Sharpening sliders, holding down the Option key ( alt key on a PC) displays B&W image of the luminance data where the sharpening will occur. This tip is important and worth remembering and doing when sharpening in ACR.

Below the Sharpening siders, is the Noise Reduction section with sliders for Luminance noise and Color noise.

The image needs to be visualized at 100% or even greater 200% or even 300% to appreciate what the noise sliders are doing. They are certainly not a replacement for Noise Ninja or NoiseWear, but they can help reduce noise.

I try to limit Luminance amounts to less than 20 or so, and usually 4 or less. Color noise reduction can be significantly higher, sometimes as high as 60 or 70. These values can help the Canon G9 images at ISO 400 or greater, which otherwise are usually not worth keeping.

This is a very brief introductory discussion of rutt's questions about Sharpening in Adobe Camera Raw. I find it very fast to do - I do it with each image on the fly, but it can be automated. If you find settings that you particularly like and use frequently, they can be stored as presets in ACR and used 'automagically'

I am most interested in hearing corrections or clarifications if I have mis-stated or mis-understood the controls in ACR sharpening.

Since doing sharpening in ACR, I find I need only mild sharpening for output and creative control. I think ACR 4.1 or better is a very significant improvement over that offered in PS CS 2.

My RAW processing of an image typically is just one or two minutes, but when finished, I receive in Photoshop a color balanced, sharp ( but not output sharpened for printing) image ready for any editing requiring the Selection Tools in a ProPhoto 16 bit color space.

xendless xurbia · April 10, 2008

For a quick look at Camera RAW sharpening (input sharpening) go to the following link and have a look halfway down the page.

http://photoshopnews.com/photoshop-cs3-for-photographers/camera-raw-41-update/

State of the art sharpening workflow?

Comments