More on sharpening in Lightroom versus Photoshop

[tonybrown]

More on sharpening in LR versus PS

I’ve been experimenting with sharpening in PS3 for a while and come up with several recipes that I like. I’m trying to convert to LR but I feel reluctance to accept the LR canned solutions that give me little insight as to what is really going on. Let me give a couple of illustrations:
The anti-aliasing filter in my 1DsIII has a recommended amount of compensating sharpening provided by Canon for the un-sharp mask, specifically, amount 240%, radius 0.4 (or less), levels 2. It has been said that in LR, sharpening is applied to the luminance channel only, and that therefore you don’t need as much – the LR maximum of 150% will do (with a radius of 0.5 pixels). I have some strong doubts that this argument holds at sub-pixel dimensions. I think that the sub pixel dimensions are intended to overcome the bandwidth limitations described by Shannon’s sampling theorem. Consequently I’ve found that a better way of  sharpening (avoiding fringe effects) is to oversample by a factor of two for an effective radius of 0.25 pixels – but that involves a file size that is four times bigger and some patience. In LR my hands are tied with no oversampling and limited sharpening. Wouldn’t it be nice to allow experienced users to evoke wider limits?
My second example relates to undesirably high contrast pictures. Typically in such cases I will adjust for linear contrast and then apply local contrast in PS with the USM set at an amount < 25%, radius 50 pixels and levels 0. This creates a picture with lots of local contrast but eliminates the extremes that are unacceptable. LR does not permit me to do this, but it does provide a clarity control that comes close – however only working on mid-tones. Again the LR canned solution provides me with creative possibilities but does not really give me a feel for what is going on behind the scenes.
For now my solution is to use LR for the nice controls it does provide but transfer to PS for better (more understandable) sharpening control. It would be nice if the PS USM could be embedded in LR.
Any thoughts?

[Josh-H]

More on sharpening in LR versus PS

I’ve been experimenting with sharpening in PS3 for a while and come up with several recipes that I like. I’m trying to convert to LR but I feel reluctance to accept the LR canned solutions that give me little insight as to what is really going on. Let me give a couple of illustrations:
The anti-aliasing filter in my 1DsIII has a recommended amount of compensating sharpening provided by Canon for the un-sharp mask, specifically, amount 240%, radius 0.4 (or less), levels 2. It has been said that in LR, sharpening is applied to the luminance channel only, and that therefore you don’t need as much – the LR maximum of 150% will do (with a radius of 0.5 pixels). I have some strong doubts that this argument holds at sub-pixel dimensions. I think that the sub pixel dimensions are intended to overcome the bandwidth limitations described by Shannon’s sampling theorem. Consequently I’ve found that a better way of  sharpening (avoiding fringe effects) is to oversample by a factor of two for an effective radius of 0.25 pixels – but that involves a file size that is four times bigger and some patience. In LR my hands are tied with no oversampling and limited sharpening. Wouldn’t it be nice to allow experienced users to evoke wider limits?
My second example relates to undesirably high contrast pictures. Typically in such cases I will adjust for linear contrast and then apply local contrast in PS with the USM set at an amount < 25%, radius 50 pixels and levels 0. This creates a picture with lots of local contrast but eliminates the extremes that are unacceptable. LR does not permit me to do this, but it does provide a clarity control that comes close – however only working on mid-tones. Again the LR canned solution provides me with creative possibilities but does not really give me a feel for what is going on behind the scenes.
For now my solution is to use LR for the nice controls it does provide but transfer to PS for better (more understandable) sharpening control. It would be nice if the PS USM could be embedded in LR.
Any thoughts?

Here is my take on it as a fellow 1DSMK3 shooter.

Firstly, its important to understand (and I believe this is correct, but some one more knowledgeable may well chime in) that the whole idea of LR is to simplify the image input output process - from shoot to finish. Not complicate it with too many options. Hence, the canned option for sharpening - which, truthfully is not really canned - at least as far as capture sharpening is concerned. LR offers a good deal of control for capture sharpening - less so for output if you use the 'canned' options - i.e. export a jpeg and just selecting the level of sharpening.

From an output perspective in LR, IF and its a big IF, you get the capture sharpening right, then all else falls perfectly into line. Going to print output and selecting standard with the appropriate media type will give optimal print sharpening as long as you got the capture sharpening correct. True.. you dont know whats going on under the hood as it where with the output sharpening, but does it really matter when the process is so simple and the results so good?

The LR sharpening algorithms are an advanced derivative of pixel genius's sharpening plug in for Photoshop - or perhaps more accurately described as the basis from which the Lightroom sharpening is based. Jeff Schewe is the best person to explain this - its his expertise that helped develop it and much of what I have said here is what I have learned from listening to Jeff in various tutorials.

CS3 is the 800 pound gorilla of the pixel editing applications - and its great for offering high level control. LR is great at what it does - making it easy, simple and quick.

As a photographer I dont really want to have to be a full on CS3 technician - I want to make photographs and process them quickly and to a very high standard. LR lets me do that - and that saves me a lot of time and time is money.

I too have experimented with sharpening 1DSMK3 files in both CS3 and Lightroom and for what its worth I feel the controls in LR offer all the capability required to get superb results - as good as is possible in CS3, but with a lot less work.

In real world terms - if the capture from the 1DSMK3 is spot on with its focus, then I find an amount of between 40 and 60 with a radius of somewhere between 0.8 and 1, with a detail setting somewhere between 25 and 40 yields optimal results for capture sharpening.

I can get the same result in CS3 - but its more work and more hassle. I dont feel CS3 offers a better result.

I have set up 6 different sharpening presets in LR for the 1DS MK3 - 3 for landscape and 3 for portrait. With one click of the mouse I can optimally capture sharpen a file. I shoot almost exclusively with the 1DSMK3 and have a lot of experience sharpening the files, and feel I have it totally nailed in LR for sharpening. I would be happy to send you my presets for your own testing if you wish.

Clarity is a separate issue - its working on mid tone contrast enhancement, giving images more punch. Its the equivalent of the technique MR wrote about using photoshop.

[Tklimek]

I've heard both Michael Reichman and Jeff Schewe indicate that they *usually* do their CAPTURE and OUTPUT sharpening in Lightroom and that *typically* there is no longer any need to use Photoshop for those two aspects of the sharpening workflow.   Lightroom at the moment does not provide for the CREATIVE sharpening aspect of the workflow.  Since Jeff Schewe has just a little something to do with the sharpening in Lightroom, I'll take his word for it.      However, your situation may be different; Jeff will forget more about sharpening than I'll ever know!  YMMV....

Cheers....

Todd in Chicago

More on sharpening in LR versus PS

I’ve been experimenting with sharpening in PS3 for a while and come up with several recipes that I like. I’m trying to convert to LR but I feel reluctance to accept the LR canned solutions that give me little insight as to what is really going on. Let me give a couple of illustrations:
The anti-aliasing filter in my 1DsIII has a recommended amount of compensating sharpening provided by Canon for the un-sharp mask, specifically, amount 240%, radius 0.4 (or less), levels 2. It has been said that in LR, sharpening is applied to the luminance channel only, and that therefore you don’t need as much – the LR maximum of 150% will do (with a radius of 0.5 pixels). I have some strong doubts that this argument holds at sub-pixel dimensions. I think that the sub pixel dimensions are intended to overcome the bandwidth limitations described by Shannon’s sampling theorem. Consequently I’ve found that a better way of  sharpening (avoiding fringe effects) is to oversample by a factor of two for an effective radius of 0.25 pixels – but that involves a file size that is four times bigger and some patience. In LR my hands are tied with no oversampling and limited sharpening. Wouldn’t it be nice to allow experienced users to evoke wider limits?
My second example relates to undesirably high contrast pictures. Typically in such cases I will adjust for linear contrast and then apply local contrast in PS with the USM set at an amount < 25%, radius 50 pixels and levels 0. This creates a picture with lots of local contrast but eliminates the extremes that are unacceptable. LR does not permit me to do this, but it does provide a clarity control that comes close – however only working on mid-tones. Again the LR canned solution provides me with creative possibilities but does not really give me a feel for what is going on behind the scenes.
For now my solution is to use LR for the nice controls it does provide but transfer to PS for better (more understandable) sharpening control. It would be nice if the PS USM could be embedded in LR.
Any thoughts?

[Josh-H]

Lightroom at the moment does not provide for the CREATIVE sharpening aspect of the workflow

Actually that isnt true.

By using the local adjustment brush and the sharpening slider in LR you can make local sharpening adjustments - or creative sharpening.

I frequently use this technique to add just a little extra sharpening to the eyes in a portrait.

[Tklimek]

Josh...

That is absolutely correct; I'm just not sure that solution is "fully qualified" for the CREATIVE portion of the workflow albeit it certainly can be used.  I think that what Jeff (not to put words in Jeff's mouth) was indicating was that the CAPTURE and OUTPUT sharpening in Lightroom is a 100% solution; I'm not too sure about the CREATIVE part.  But, to be honest, I'm not an expert.  I also use the TAT brush to sharpen some areas (same as you such as eyes in a portrait).

Cheers...

Todd in Chicago

Actually that isnt true.

By using the local adjustment brush and the sharpening slider in LR you can make local sharpening adjustments - or creative sharpening.

I frequently use this technique to add just a little extra sharpening to the eyes in a portrait.

[jpegman]

One other major issue is what version of LR you are comparing it to?

LR2 is vastly improved over LR1 with ACR capture sharpening, and the output sharpening has been also improved to the point at Photoshop World in Boston last week, I would guess that 80-90% of the instructors in the classes I attended said that they use CS4 for pixel level editing when needed and softproofing only - everything else is done through Lightroom 2.3

As Josh H indicated, LR2 with the local adjustment brush technology can do localized sharpening (or color, brightness, etc) so LR2 has improved across the board from LR1, and some at PSW hinted that softproofing may show up in LR3.

Jpegman

[madmanchan]

Capture and output sharpening in LR have already gone through a few development iterations and hence we feel they're tuned fairly well. The creative sharpening in LR, accessed via local adjustments, has really only gone through a single dev cycle (i.e., we did the initial implementation, and that was it). While all 3 will improve in the future, you can make an educated guess as to which of these will receive the most immediate attention.

Tony, I don't understand what you mean by "subpixel dimensions." Your image is discretely sampled at capture time and any pixels you introduce by upsampling prior to sharpening are simply interpolated (or, if using a fancier algorithm, in computer vision we might say "hallucinated") from the existing ones, hence not really providing any additional information -- at least not from an information theory point of view. I also strongly doubt that you really need sharpening beyond 150% amount; if you find that you do, then I humbly suggest that you are misunderstanding the sharpening controls, or possibly that your original image is grossly misfocused.

My opinion is that while it is interesting (from an academic and intellectual point of view) to know what is going on behind the scenes with these controls, you can still get excellent quality results even if you don't know. I say this not because I want to sweep the details under the rug, but simply because it's what I've found working with both pro and amateur photographers (many of whom don't know and/or don't care about the underlying math) and seeing what results they come up with. The key is to learn the visual effect of the control, recognize its limitations, and understand possible interactions with other controls (for example, all the 4 capture sharpening controls are intended to be used together, not individually). You can certainly get pixel-level editing capabilities in PS for the ultimate control, but I am not at all convinced that it automatically means superior results for photographic capture sharpening.

[pindman]

OT - Sorry

Eric,

Now all we need to do is get a real curves adjustment in LR instead of the crippled parametric curves.  Just something to take back to Adobe.

Paul

[tonybrown]

Capture and output sharpening in LR have already gone through a few development iterations and hence we feel they're tuned fairly well. The creative sharpening in LR, accessed via local adjustments, has really only gone through a single dev cycle (i.e., we did the initial implementation, and that was it). While all 3 will improve in the future, you can make an educated guess as to which of these will receive the most immediate attention.

Tony, I don't understand what you mean by "subpixel dimensions." Your image is discretely sampled at capture time and any pixels you introduce by upsampling prior to sharpening are simply interpolated (or, if using a fancier algorithm, in computer vision we might say "hallucinated") from the existing ones, hence not really providing any additional information -- at least not from an information theory point of view. I also strongly doubt that you really need sharpening beyond 150% amount; if you find that you do, then I humbly suggest that you are misunderstanding the sharpening controls, or possibly that your original image is grossly misfocused.

My opinion is that while it is interesting (from an academic and intellectual point of view) to know what is going on behind the scenes with these controls, you can still get excellent quality results even if you don't know. I say this not because I want to sweep the details under the rug, but simply because it's what I've found working with both pro and amateur photographers (many of whom don't know and/or don't care about the underlying math) and seeing what results they come up with. The key is to learn the visual effect of the control, recognize its limitations, and understand possible interactions with other controls (for example, all the 4 capture sharpening controls are intended to be used together, not individually). You can certainly get pixel-level editing capabilities in PS for the ultimate control, but I am not at all convinced that it automatically means superior results for photographic capture sharpening.

‘madmanchan’ let me put my sub-pixel thoughts in context. First, I don’t question that most LR users find it gives them great and rapid control of their photographs and in most cases their results are really pleasing. No doubt this is why the LR limits on sharpening are maximum amount of 150% and smallest radius of 0.5 pixels – it’s a good compromise for most people’s purposes. PS lets you go way beyond these limits and Canon have suggested on their site that the anti-aliasing filter is best compensated by sharpening at 240% and 0.4 pixel radius maximum. In practice this gives a nice sharpening effect without any significant fringing. Judged by Canon’s recommendation, Lightroom seems to be limiting.
My comment on sub-pixel sharpening was anecdotal in nature, as I’m an engineer with a speciality in communications and semiconductors. I’ve experimented with sharpening by oversampling the image and it does give you more control over the shape of the sharpening curve – you can get good results with a lower amount of sharpening. Its interesting, but requires lots more computing and for most people is going in the wrong direction.
Most posts in this thread didn’t address the other example I gave, of creating improved local contrast with USM set to amount of 25% (max) and radius of 50 pixels. This you simply cannot do in LR. This technique comes in useful when you photograph a winter scene at noon with a bright winter sun and deep blue cloudless sky – creating a very wide dynamic range. This needs to be developed with linear contrast which makes a rather flat image. Contrast can then be effectively restored locally using the USM as given. Michael Reichmann wrote this up on Luminous Landscape, titled ‘Understanding Local Contrast Enhancement’.
For your interest: http://photo.net/photodb/photo?photo_id=8295125

[PeterAit]

OT - Sorry

Eric,

Now all we need to do is get a real curves adjustment in LR instead of the crippled parametric curves.  Just something to take back to Adobe.

Paul

You have to wonder how much Adobe is going to put in LR. Too much, and people won't need Photoshop any more! Bad for business.

Peter

[Misirlou]

This is an astonishingly valuable and coherent thread. Are we still on the Internet?

[digitaldog]

Now all we need to do is get a real curves adjustment in LR instead of the crippled parametric curves.  Just something to take back to Adobe.

Well if someone can demonstrate that its necessary, as well as proving to the team that the current curves are "crippled" you might get that wish.

One problem is, folks think way too much of curves, used them for so many years in Photoshop that they instantly migrate to the curves in LR or ACR which is kind of a mistake. I find I rarely need to use Curves there and when I do, its for minor tweaks. I'm not anti-sliders either.

That said, a saturation curve in both the ACR engine and in Photoshop would be real useful in a few rare situations.

[ErikKaffehr]

Hi,

My two cents on this is that "Capture" and "Output" sharpening are pretty optimal in Lightroom. Regarding "creative" sharpening there are a lot of tools in Photokit Sharpener. I do apply sharpening on layers with mask quite often. Just as an example, Photokit Sharpener has a "depth of field" plugin which can fake some DOF. This of course needs to be applied selectively. Normally I would use a gradient mask. A similar situation may arise when using more extreme wide angles which often are quite soft in the corners, in this case I can do agressive sharpening on a layer and use a circular gradient as mask so it will be applied on corners/edges but not in the center.

Regarding capture sharpening I'd suggest that Lightroom is quite OK. Reinventing Lightroom capture sharpening with unsharp mask would be hard work. The reason is that Lightroom presharpening has some more features than "unsharp mask", there are options for gradient based masks and halo avoidance in addition to radius and amount. AFAIK the amount and radius sliders may not have a 1:1 correspondance in LR and Photoshop "unsharp mask".

Best regards
Erik

Josh...

That is absolutely correct; I'm just not sure that solution is "fully qualified" for the CREATIVE portion of the workflow albeit it certainly can be used.  I think that what Jeff (not to put words in Jeff's mouth) was indicating was that the CAPTURE and OUTPUT sharpening in Lightroom is a 100% solution; I'm not too sure about the CREATIVE part.  But, to be honest, I'm not an expert.  I also use the TAT brush to sharpen some areas (same as you such as eyes in a portrait).

Cheers...

Todd in Chicago

[Josh-H]

AFAIK the amount and radius sliders may not have a 1:1 correspondance in LR and Photoshop "unsharp mask".

That is a really important point and one I wanted to make earlier - but I couldnt remember where I had read / heard that, nor could I remember what the actual relationship was.

I remember Jeff Schewe talking about the relationship not being 1:1 - but for the life of me cant remember where this was or any of the details.

I do vaguely remember that a setting of .8 in LR was approximately equivalent to a setting of .5 in Photoshop USM.

would love it if someone could clarify.

[ErikKaffehr]

Yes, that's true to an extent. On the other hand if Lightroom is not good enough users just migrate to Aperture.

Keep in mind that Lightroom is a parametric editor, it will never will be able to do pixel level manipulation except for very limited tasks. So for anything related to pixels you need Photoshop or something similar. IMHO customers doing serious business will need Photoshop, because that is the industry standard. It does cost like 1000 USD but that is pretty much ignorable compared to costs for other equipment, even if they of course add up. You need a "color calibration device", a decent screen and so on.

Some folks may do with simpler tools, like Photoshop Elements. There are a lot of hidden functions in Elements that can be enabled with "Elements+", see link below:
http://simplephotoshop.com/elementsplus/index.htm
Best regards
Erik

You have to wonder how much Adobe is going to put in LR. Too much, and people won't need Photoshop any more! Bad for business.

Peter

[Tklimek]

Absolutely correct.....the one thing that is mentioned for certain is that the radius amount is not a 1:1 with Photoshop USM.  My guess is that the other sliders are not either, but maybe when Jeff comes back from his trip he can pipe in on this thread (unless Eric knows...).

Jeff mentions this in the Lightroom 2 tutorial with Michael, chapter 13 "Develop_Detail".

Cheers...

Todd in Chicago

That is a really important point and one I wanted to make earlier - but I couldnt remember where I had read / heard that, nor could I remember what the actual relationship was.

I remember Jeff Schewe talking about the relationship not being 1:1 - but for the life of me cant remember where this was or any of the details.

I do vaguely remember that a setting of .8 in LR was approximately equivalent to a setting of .5 in Photoshop USM.

would love it if someone could clarify.

[Ernst Dinkla]

Yes, that's true to an extent. On the other hand if Lightroom is not good enough users just migrate to Aperture.

Best regards
Erik

It is of course almost blasphemy but could it be that Adobe also wants to compete with a small company like DDIsoftware aka Qimage?  There still are Q output features not matched by Lightroom. Preventing migration to Aperture must be much easier than attracting immigrants from Qimage.


met vriendelijke groeten, Ernst Dinkla

Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/

[NikoJorj]

One problem is, folks think way too much of curves, used them for so many years in Photoshop that they instantly migrate to the curves in LR or ACR which is kind of a mistake.

That took me a while to get used to it, but I'll just second that for global image adjustments : the possible oddities of a true point curve (ie past the ones you can produce with a parametric curve) just don't make much sense in a photographic context.

However, there is one area where "true" (kinked and skewed) point curves might still make sense : local image adjustments.
When just a part of the image is selected, the corresponding histogram might cover just a part of the 0-100 range and then, "unusual" curves could make sense.
I'll have an example at hand this evening, where I'd like to bring down a bit the highlights on a small clouds area, without affecting the midtones (ie shadows, rel. speaking) of these clouds (and obviously without dulling the rest of the highlights with a global recovery).

[tonybrown]

That is a really important point and one I wanted to make earlier - but I couldnt remember where I had read / heard that, nor could I remember what the actual relationship was.

I remember Jeff Schewe talking about the relationship not being 1:1 - but for the life of me cant remember where this was or any of the details.

I do vaguely remember that a setting of .8 in LR was approximately equivalent to a setting of .5 in Photoshop USM.

would love it if someone could clarify.

Finally! I've found my reference to the Canon advice on sharpening for the EOS 1Ds Mark III (and 1D Mark III). Its in a document titled "Tayloring the camera for different situations" and on page 16 it reads:

Suggested starting points for Unsharp Mask sharpening in the computer
For shooters who anticipate making further sharpening adjustments in Adobe Photoshop™ or similar software programs,
always bear in mind that ideal sharpening settings will be very much dependent upon subject matter, lighting, ISO setting used, and of course the intended type and size of final output. As a typical, initial starting point for sharpening using Photoshop’s Unsharp Mask command, you might want to consider the following :
High ISO images (no initial in-camera sharpening)
Amount — 250%, Radius — 0.3 pixels, Threshold — 4 levels
Low ISO images (no initial in-camera sharpening)
Amount — 250%, Radius — 0.3 pixels, Threshold — 1 levels


So Josh: following your suggestion that 0.8 in LR is roughly equivalent to a setting of 0.5 in Photoshop USM, Canon's recommendation of Radius = 0.3 pixels might look like 0.5 pixels in LR, which is doable. However, the amount = 250% in PS USM is quite a bit more than one can do in LR, unless its possible to do it in two steps? Somewhere I read that Canon's antialias filter is quite conservative, i.e needs more sharpening in the develop phase, which may explain the high 250%. The web reference for the Canon paper is:
http://www.usa.canon.com/uploadedimages/FC...tings_Final.pdf

 Regards Tony

[tonybrown]

I have set up 6 different sharpening presets in LR for the 1DS MK3 - 3 for landscape and 3 for portrait. With one click of the mouse I can optimally capture sharpen a file. I shoot almost exclusively with the 1DSMK3 and have a lot of experience sharpening the files, and feel I have it totally nailed in LR for sharpening. I would be happy to send you my presets for your own testing if you wish.

Josh: I'd like to take you up on your offer of the sharpening presets you have set up. Do you need my email address?
Also, I'd like to say I appreciate the way this thread has been handled by everyone.
Regards, Tony