ACR 4.4 and Clarity

[DarkPenguin]

CLARITY HI-JACKED!
[a href=\"index.php?act=findpost&pid=183825\"][{POST_SNAPBACK}][/a]

One has to ask, did this thread provide any clarity?

[Mark D Segal]

One has to ask, did this thread provide any clarity?
[a href=\"index.php?act=findpost&pid=183922\"][{POST_SNAPBACK}][/a]

As a participant who tried, I shall have to recuse myself from trying to answer that (good) question!  

[Philmar]

hmmm, is it safe to wade back in and discuss 'clarity'?

ACR 4.4(and the LR version) have been pulled and is being recommended that you should roll back to v 4.3x. Noise reduction should be done before sharpening if possible. Why would you want to sharpen noise?
[a href=\"index.php?act=findpost&pid=183631\"][{POST_SNAPBACK}][/a]

What if you prefer to apply NR to a image with a layer mask?
So do you use the noise reduction in ACR?  Isn't the ACR noise reduction as crude as the ACR sharpening once was, precipitating the use of NR outside of the converter (as it was once better to do capture sharpoening outside of ACR)?

And I, for one, would also appreciate and benefit from any small tutorial on how to use the clarity slider. Rarely do I set it much beyond '10', except on photos with smog or other air particles. I've read somewhere that you need to judge it's application when viewed at 100%. I find this really doesn't help much. I see how when judging sharpening that 100% viewing helps. I can see the haloes better. But when viewing the effects of moving the clarity slider at 100% view I find it really doesn't help me.

[DarkPenguin]

Not addressed to me but I'll take a stab.

What if you prefer to apply NR to a image with a layer mask?

Then doing it in the converter isn't possible and you have to do it later.

So do you use the noise reduction in ACR?  Isn't the ACR noise reduction as crude as the ACR sharpening once was, precipitating the use of NR outside of the converter (as it was once better to do capture sharpoening outside of ACR)?
[a href=\"index.php?act=findpost&pid=183946\"][{POST_SNAPBACK}][/a]

I've found that for most of my cameras after my 300D (20D, Xti and 40D) ACR's noise reduction is just fine.  For those images where it is not (read:G9) I turn off sharpening in ACR and do noise reduction and sharpening in PS.

[digitaldog]

I've found that for most of my cameras after my 300D (20D, Xti and 40D) ACR's noise reduction is just fine.  For those images where it is not (read:G9) I turn off sharpening in ACR and do noise reduction and sharpening in PS.
[a href=\"index.php?act=findpost&pid=183948\"][{POST_SNAPBACK}][/a]

With a 5D, I'd agree too. Now, if and when I shoot 3200, I'm getting better results using Noiseware in Photoshop but that's not something I would need for all images. Jeff might pipe in here, there was a pretty big increase in quality of noise reduction between version 1.0 and one of the later builds (1.3?). The defaults settings when viewed as they should be (100%) do help and of course you can alter those settings.

[Panopeeper]

if and when I shoot 3200, I'm getting better results using Noiseware in Photoshop

This is surprizing, for ISO 3200 is a numerical derivative of ISO 1600 on the 5D, so the noise structure must be identical.

[Ray]

Hi Ray, that's a very interesting pair of images. I think what it shows is that while RSP bundled local contrast enhancement and sharpening in one very effective tool, with ACR you would need to use two tools for roughly equivalent effect. Your left image shows pretty much what is to be expected from ACR - at zero Amount you get zero sharpening regardless of where you set Detail.

[a href=\"index.php?act=findpost&pid=183890\"][{POST_SNAPBACK}][/a]

Hi Mark,
I take your point that 'detail' in ACR has no effect unless one also applies some sharpening. With sharpening at zero, it's irrelevant what setting 'detail' is on.

However, having applied 50% sharpening with 'detail' at 100%, I get a similar effect to RSP's 'detail extraction' at maximum (no sharpening), except there's an appealing quality of color solidity to the RSP image which I fail to emulate in ACR.

Perhaps you, or Andrew Rodney or Jeff Schewe would have no trouble emulating this RSP effect in ACR. I cannot.

[attachment=5726:attachment]

[Mark D Segal]

Hi Mark,
I take your point that 'detail' in ACR has no effect unless one also applies some sharpening. With sharpening at zero, it's irrelevant what setting 'detail' is on.

However, having applied 50% sharpening with 'detail' at 100%, I get a similar effect to RSP's 'detail extraction' at maximum (no sharpening), except there's an appealing quality of color solidity to the RSP image which I fail to emulate in ACR.

Perhaps you, or Andrew Rodney or Jeff Schewe would have no trouble emulating this RSP effect in ACR. I cannot.

[attachment=5726:attachment]
[a href=\"index.php?act=findpost&pid=184063\"][{POST_SNAPBACK}][/a]

Ray, I'm not too sure what "color solidity" is, but I'll surmise from looking at your illustration that it is a combination of saturation and clarity. I agree your RSP example looks quite smart. Looking at that image, have a go at setting ACR's Recovery to a value somewhere between 3~5 to tame the highlights which look close to blowing. If you are not yet at clipping in the blacks, I would then slide the Blacks up a value of 1 or 2. This tends to add "body" to all colours as well as to improve contrast. Now, if you think there is not quite enough tonal separation in the dark range of the image, increase Fill - I would suggest to no more than a value of 20, then increase Blacks again to the clipping point. Then I would boost both Vibrance and Clarity to taste, and have a second look at the sharpening settings - particularly the Radius - setting it just a wee bit to the right in the context of those other adjustments could take you very close to equilibrating the two images or even better. Then you have a whole panoply of other controls in Curves and HSL if it hasn't yet floated your boat................

[digitaldog]

This is surprizing, for ISO 3200 is a numerical derivative of ISO 1600 on the 5D, so the noise structure must be identical.
[a href=\"index.php?act=findpost&pid=183991\"][{POST_SNAPBACK}][/a]

When I shoot ISO 1600 I still get better results using Noiseware in Photoshop....

[Philmar]

I guess what I am trying to ascertain is whether it is better to use ACR's (what I thought were) crude NR capabilities or whether it is better to use a 3rd NR dedictaed plugin (like Ninja, Neat Image ect.).

I understand that theoretically it is best to do NR early as possible, so that one doesn't sharpen the noise before the NR is applied. Obviously it can't be done any earlier than during RAW conversion.
However I also realise that NR, like sharpenig, is best applied using a layer mask so that only the areas with noise are targeted. This isn't possible in ACR (unless it too uses the same mask, the inversion, that can be used for capture sharpening).
So what is the better route?
a) Early in ACR which applied globally would soften non-noisy areas

OR

 in PS employing a layer mask that targets the noisy surface regions.

[NikoJorj]

I guess what I am trying to ascertain is whether it is better to use ACR's (what I thought were) crude NR capabilities or whether it is better to use a 3rd NR dedictaed plugin (like Ninja, Neat Image ect.).
[...]
So what is the better route?
a) Early in ACR which applied globally would soften non-noisy areas
[a href=\"index.php?act=findpost&pid=184530\"][{POST_SNAPBACK}][/a]

My 2c's : I'd vote for this one - as long as we're not talking of a 1/3" digicam @6400ISO  , the softening will be rather stealth.
There will still be some luminance noise revealed after vigorous contrast enhancements, which may not be that objectionable after all as long long there is enough detail underneath...

It's a matter of taste, definitely, as there is always a tradeoff between noise and detail.

[Philmar]

My 2c's : I'd vote for this one - as long as we're not talking of a 1/3" digicam @6400ISO  , the softening will be rather stealth.
There will still be some luminance noise revealed after vigorous contrast enhancements, which may not be that objectionable after all as long long there is enough detail underneath...

It's a matter of taste, definitely, as there is always a tradeoff between noise and detail.
[a href=\"index.php?act=findpost&pid=184547\"][{POST_SNAPBACK}][/a]

Thanks for your reply. I am curious how many of the pros here resolve this dilemma.

[eronald]

Andrew,
 I'm entertained by your comment about proving things scientifically, but simple reasoning should suffice for our little topic here - it's not rocket science. - As for existing marketed methods not working - I agree

1. There is no color directly associated with the pixel data in a Raw file. The color is (re)created as part of the Raw conversion process, which then maps the pixels into a color space.

2. We call this mapping process "rendering". It relies on information about the camera to create colors. Eg. identical Bayer pixel data from a different Canon or Nikon model will render to a different color file.

3. Therefore some external information other than pixel data is key to establishing the color of the rendering. This information is camera-dependent.

4. Hence, unless the software manufacturer is omniscient, better information about the camera used to make the capture —calibration information— will improve the quality of the rendering.

Edmund

So again, with or without the script, the ONLY way you can render an image to a color appearance you so desire is to use a package that allows you to use a custom camera profile? If so, some of us want to examine your Raw files and what you're attempting to render.

You may have had conversations with Thomas, but it appears like so many before you, your desire for a function in an Adobe product was either ill defined, is a solution is search of a problem or, you don't have the problem you think you do.

Lots of Adobe users ask for feature requests, so few know how to do so correctly and based on the very smart people at Adobe who decide how to design a product. You need to demonstrate scientifically (or supply the math, good luck) that without the ability to build and use custom ICC profiles in ACR, the product is unable to produce the results most users require. The group grumbling about this are often those who sell profiles, or software to build profiles. That's not at all been a convincing group so far, probably because they have totally failed to illustrate other than a want, why this is necessary. Thomas didn't leave out ICC profiles to piss off companies trying to sell ineffective camera profiling software, it did it because none of these people have illustrated why its at all useful to anyone but those companies.
[a href=\"index.php?act=findpost&pid=183731\"][{POST_SNAPBACK}][/a]

[digitaldog]

2. We call this mapping process "rendering". It relies on information about the camera to create colors. Eg. identical Bayer pixel data from a different Canon or Nikon model will render to a different color file.

Here's where we go in different directions. Rendering (the process of conversion from scene to output referred) is the job of the photographer or image creator using the Raw converter. This isn't an automatic process (otherwise, lets just shoot JPEGs from in camera processing). And as yet, the pro ICC camp has not proved that using a profile automatically produces the desired color appearance for all scenes forging the need to use the supplied rendering controls in a converter.

No one in the pro ICC custom camera profile camp has suggested that a profile is a magic bullet that automatically by default makes this rendering preferred, correct or otherwise because that's simply not what happens. They might suggest "It provides a better starting point" (an argument that those who shoot Raw with a custom white balance suggest), but in both cases, white card or ICC profile, no one has or should (or can) demonstrate that using either produces the desired color without user intervention. That being the case, then I turn the question the other way: without either technique, are you unable to produce a desired color appearance with the current toolset. As yet, the pro ICC camp refuses to answer this. Or, if the default rendering takes 8 steps to get to a closer starting point, why not just do this once, save a preset? Its going to be a lot faster and easier than trying to photograph a target to build a profile, let alone build the profile too.

Basically, the side that says "We must have access to profiles" has totally failed to express why they must have them, they've totally failed to prove that the current products discussed are unable to produce a desired color appearance without said profiles. They've totally failed to prove that colorimetrically correct capture is anything but a starting point for rendering by the user. They've totally failed to prove to the pretty smart color guys at Adobe that such profiles are necessary.

[eronald]

Andrew,

A profile or matrix is there already in the Adobe software; this is like a canned printer profile -good but can be improved. And this matrix sets the place where automation stops and the photographer's job starts.


If you wish, the Adobe profile is the equivalent of the native Epson printer driver - it gives excellent results, but someone with access to a spectro and software can improve the print channel or customize the rendering of a print to specific light

In the same way, someone with access to a spectroradiometer to measure ambient light, and a monochromator to measure the camera's spectral response can significantly tailor the quality of the Raw rendering.

Of course, this just sets a better starting point for the photographers's work. Andthis is nothing like the profiling packages now marketed.

Last but not least, Andrew, you keep saying Adobe, Adobe, Adobe, but in fact the camera guys make the cameras, not Adobe, in the same way the printer guys make the printers, not Adobe.  And the camera guys seem to find the idea of customized camera matrices quite reasonable - you think they don't understand their equipment ?

Edmund

Basically, the side that says "We must have access to profiles" has totally failed to express why they must have them, they've totally failed to prove that the current products discussed are unable to produce a desired color appearance without said profiles. They've totally failed to prove that colorimetrically correct capture is anything but a starting point for rendering by the user. They've totally failed to prove to the pretty smart color guys at Adobe that such profiles are necessary.
[a href=\"index.php?act=findpost&pid=184705\"][{POST_SNAPBACK}][/a]

[digitaldog]

A profile or matrix is there already in the Adobe software; this is like a canned printer profile -good but can be improved.

Well other than using the Calibrate tab, that's yet to be proven. Even without the calibrate tab, plenty of users provide a color rendering they desire. So this goes back to the question of, can you or can you not produce the color you desire and is using the Calibrate tab OR some custom profile going to do much for you. I suspect the vast majority of LR/ACR users are not messing around with Calibrate. In fact while I did run this routine on my previous camera, I've found no reason to do so with my 5D. Nor do I have an issues getting out of LR what I want.

And this matrix sets the place where automation stops and the photographer's job starts.
If you wish, the Adobe profile is the equivalent of the native Epson printer driver - it gives excellent results, but someone with access to a spectro and software can improve the print channel or customize the rendering of a print to specific light.

A far better analogy would be, here's where the Epson driver starts WITH a canned Epson profile versus building your own. With just the Epson driver, yes I can get a decent print. Without either the canned or custom profile, I can't soft proof, I can't get the RGB values IN Photoshop to send to the printer etc. The profile(s) we have in ACR/LR are where we now stand with good canned profiles from Epson. You might be able to get slightly better results rolling your own but the vast majority of users don't need to. And, unlike printing, Raw conversion is about moving all kinds of sliders around to get a desired color appearance. You do that before you print, then you want "what you see to be what you get". In this context, we're talking about getting what you want visually from the Raw data.

In the same way, someone with access to a spectroradiometer to measure ambient light, and a monochromator to measure the camera's spectral response can significantly tailor the quality of the Raw rendering.

That would get us far closer to this desired goal of a scene built input profile yes. We're not there now. The current set of solutions is based on the false idea that if you shoot some target (with a gamut that's a far cry from scene gamut), that the dynamic range of the scene and target are the same, you someone fingerprint the way the device captures the color which is still scene referred and has to be output referred. At least with the above idea, you can measure the illuminate and items in the scene if so desired. And the profile here is capture specific, not a single (or better in the case of LR/ACR duel) profile that is supposed to fingerprint anything in front of the sensor.

Even still, its doesn't (nor will it) guarantee the user will not need to alter the rendering controls to get what they desire. No more than shooting a Macbeth with 4 different kinds of E6 film will produce either the same color appearance or the "correct" color appearance.

[Mark D Segal]

Furthermore, to improve upon the canned math in ACR you would need two necessary conditions: (a) you can do a better job profiling the camera than Thomas Knoll can, and ( there is noticeable copy-to-copy variability in the capture characterestics of the camera. My guess is that you would be very hard pressed on (a), and for ( - it doesn't apply - at least to the higher-end pro DSLRs, such as 1Ds series, 5D, D3, etc. I think that's why the majority of users, me includeed, ignore the calibrate tab.

What I would REALLY like, however, is some firmware in Canon cameras which allows users to find a systematic way of getting them to do ETTR instead of ETTL!

[Mark D Segal]

I guess what I am trying to ascertain is whether it is better to use ACR's (what I thought were) crude NR capabilities or whether it is better to use a 3rd NR dedictaed plugin (like Ninja, Neat Image ect.).

I understand that theoretically it is best to do NR early as possible, so that one doesn't sharpen the noise before the NR is applied. Obviously it can't be done any earlier than during RAW conversion.
However I also realise that NR, like sharpenig, is best applied using a layer mask so that only the areas with noise are targeted. This isn't possible in ACR (unless it too uses the same mask, the inversion, that can be used for capture sharpening).
So what is the better route?
a) Early in ACR which applied globally would soften non-noisy areas

OR

 in PS employing a layer mask that targets the noisy surface regions.
[a href=\"index.php?act=findpost&pid=184530\"][{POST_SNAPBACK}][/a]

The mask in the Detail tab of Camera Raw applies for sharpening, not for noise reduction. If there is noise you wish to reduce - but not in ACR, then you should not sharpen in ACR either. Like you, I often prefer to target noise reduction to those areas where it shows, which needs layer masking that cannot be done in ACR. If you try noise reduction in ACR and find it satisfactory, then capture sharpening in ACR also makes sense.

[eronald]

Furthermore, to improve upon the canned math in ACR you would need two necessary conditions: (a) you can do a better job profiling the camera than Thomas Knoll can, and ( there is noticeable copy-to-copy variability in the capture characterestics of the camera. My guess is that you would be very hard pressed on (a), and for ( - it doesn't apply - at least to the higher-end pro DSLRs, such as 1Ds series, 5D, D3, etc. I think that's why the majority of users, me includeed, ignore the calibrate tab.

What I would REALLY like, however, is some firmware in Canon cameras which allows users to find a systematic way of getting them to do ETTR instead of ETTL!
[a href=\"index.php?act=findpost&pid=184725\"][{POST_SNAPBACK}][/a]

 It's pretty obvious that I can do a better job of calibrating *my* camera in my hand than Thomas Knoll can without seeing it, ever, if I have access to lab grade equipment to spectrally measure my camera's sensor response, and the ability to spectrally measure the lighting under which I take my shot.

 In the same way,  seven or eight years ago Andrew could certainly make a better profile of *his own* printer than the canned one supplied by a manufacturer. Of course, back in the days when he started making profiles the equipment (spectrophotometer) he needed was pretty rare and expensive, and could almost be considered lab grade. Nowadays,the printers are matched to close tolerances, and maybe even Andrew is the consultant who makes those canned profiles, so I guess then the  canned profiles match his own printers


 I'm not speaking off the top of my head here - I discuss this stuff with the camera manufacturer's techs and color scientists and Raw processor writers.

Edmund

[digitaldog]

It's pretty obvious that I can do a better job of calibrating *my* camera in my hand than Thomas Knoll can without seeing it, ever, if I have access to lab grade equipment to spectrally measure my camera's sensor response, and the ability to spectrally measure the lighting under which I take my shot.

You don't need lab grade devices to do this. And what makes you think this is how Thomas makes his profiles in the first place (or that having such a device is necessary)?

Nowadays,the printers are matched to close tolerances, and maybe even Andrew is the consultant who makes those canned profiles, so I guess then the  canned profiles match his own printers

Few could see the difference between the custom profile I built for Exhibition Fiber paper for my 3800/4800/7880 and the canned profiles I built for Epson on the same machines.

I'm not speaking off the top of my head here - I discuss this stuff with the camera manufacturer's techs and color scientists and Raw processor writers.

Well most of them (however they are) have done a piss poor job of convincing anyone else that its necessary, at least some of us color geeks and those at Adobe. Of course I've had no success speaking with camera manufactures techs about getting them to provide something reasonably useful in terms of the Raw data info on the back of their cameras.