Ad
Ad
Ad
Pages: « 1 ... 5 6 [7] 8 »   Bottom of Page
Print
Author Topic: ACR 4.4 and Clarity  (Read 70507 times)
picnic
Sr. Member
****
Offline Offline

Posts: 574


« Reply #120 on: April 12, 2008, 05:13:03 PM »
ReplyReply

Quote
Good news for you: DPP DOES have CA control. Moreover, it has controls for vignetting correction (like ACR), distortion correction (unlike ACR), and some more - all this with the knowledge of the lens.

[a href=\"index.php?act=findpost&pid=188823\"][{POST_SNAPBACK}][/a]

Unfortunately that is only Canon lenses--and not all of them.  

Diane
Logged
eronald
Sr. Member
****
Offline Offline

Posts: 3626



WWW
« Reply #121 on: April 12, 2008, 05:28:34 PM »
ReplyReply

Quote
I'm on that board in some capacity so I have some idea how things move. I wish them all luck and would add, all their efforts are appreciated. But so far, there's a big solution in search of a problem I've yet to see anyone on your side define. Until that happens, or the technology improves to the degree we can actually measure the scene illuminant, the current solutions are half baked.
[a href=\"index.php?act=findpost&pid=188981\"][{POST_SNAPBACK}][/a]

Andrew

All of us techies on that board have started to have a hard look at the issues with Raw lately. The general feeling seems to be that the time to fully color-manage Raw has come. The problem is now hacking out a solution that keeps all the proprietary interests happy.

AFAIK The capacity to spectrally measure the scene illuminant is there, fully implemented, in every EyeOne Pro and every sub $500 ColorMunki. Of course at this point you need a computer set up in the studio to do the incident light measure, but that's no problem for the tethered studio shooters. Of course at this point, there is no packaged software available that would allow a photographer to use those spectra, but as a color geek, if if you also have the sensor response you can do anything you want with the camera, including computing the Knoll matrices for the measured illuminant

The problem at that point becomes convincing LR to use *your* Knoll matrix rather than its internal preset Knoll matrix. And of course the fact that the camera color geeks say that a Knoll matrix will not solve everything because of camera metamerism ...
 
Edmund
« Last Edit: April 12, 2008, 05:42:28 PM by eronald » Logged

Edmund Ronald, Ph.D. 
digitaldog
Sr. Member
****
Offline Offline

Posts: 8027



WWW
« Reply #122 on: April 12, 2008, 05:54:30 PM »
ReplyReply

Quote
Andrew
All of us techies on that board have started to have a hard look at the issues with Raw lately. The general feeling seems to be that the time to fully color-manage Raw has come.
[a href=\"index.php?act=findpost&pid=189070\"][{POST_SNAPBACK}][/a]

What currently isn't color managed with respect to Raw?

Quote
The problem is now hacking out a solution that keeps all the proprietary interests happy.

That sounds like a political issue and I'm not at all interested in having anything to do with politics and digital imaging. I'm interested in what problems we have and how they can be corrected from a technical and end user perspective.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
eronald
Sr. Member
****
Offline Offline

Posts: 3626



WWW
« Reply #123 on: April 12, 2008, 06:24:44 PM »
ReplyReply

Quote
What currently isn't color managed with respect to Raw?
That sounds like a political issue and I'm not at all interested in having anything to do with politics and digital imaging. I'm interested in what problems we have and how they can be corrected from a technical and end user perspective.
[a href=\"index.php?act=findpost&pid=189075\"][{POST_SNAPBACK}][/a]

Andrew,

At the moment, nothing about Raw is color managed - in the sense that every converter comes with its own cam2xyz interpretation of the Raw data into color (xyz) data which is then rendered into some output-referred space. Each converter is basically a black box with its own non-swappable cam2xyz: Canon has DPP, Nikon has NikonView, Adobe has ACR etc - none of these have any reason to associate the same xyz to a given cam pixel Raw value.

There is also  some debate at the ICC about what one should do with the demosaiced data: In which color-space one can and should encode the demosaiced scene-referred *colorimetric* data from cameras - I'm sure you're as much aware of the problems here as I am.

I'm talking to the camera guys, and there appear to be some technical subtleties involved in the cam2xyz transform; I think photographers would benefit if people from the hardware and software sides of the business were talking to each other.

Edmund
« Last Edit: April 12, 2008, 06:38:08 PM by eronald » Logged

Edmund Ronald, Ph.D. 
digitaldog
Sr. Member
****
Offline Offline

Posts: 8027



WWW
« Reply #124 on: April 12, 2008, 06:37:33 PM »
ReplyReply

Quote
At the moment, nothing about Raw is color managed

Raw has to be output referred and rendered. The color numbers I see, the color appearance I see is honored from start to finish. Raw or color neg, converter or color darkroom (or even scanning UI), nothing useful in terms of color management is necessary for the user before they even render an initial preview. One has to render a desired appearance. That's why we shoot and use Raw. From that point on, everything is fine, and its questionable in a big way IF prior to the rendering, the user has any need of color management.

We have color neg auto analyzers in labs which can do a pretty good job, much of the time, rendering a print from a color neg. That doesn't diminish the other people who need custom prints made and often find, they themselves have to control the rendering. The auto button in LR works and fails, but expecting 90%+ preferred rendering (considering how damn subjective it is), is a pretty huge stretch. And even if its 50%, it only takes a few seconds to move a few sliders to produce far more than 90% desired global rendering. My god, this workflow is really fast. Its not like we're trying to force edits on 12 million pixels in memory.

Its funny, you give people a true digital darkroom, they want it to roughly behave like the in-camera JPEG rendering engine they are trying to get away from.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
eronald
Sr. Member
****
Offline Offline

Posts: 3626



WWW
« Reply #125 on: April 12, 2008, 06:41:14 PM »
ReplyReply

Quote
Raw has to be output referred and rendered.
[a href=\"index.php?act=findpost&pid=189079\"][{POST_SNAPBACK}][/a]


I differ. Raw needs to be taken to a large space eg xyz; then it can be rendered to an output referred space. Or just written out. With people looking at things on HDR displays, the meaning of rendering is going to change.

Edmund
« Last Edit: April 12, 2008, 06:42:50 PM by eronald » Logged

Edmund Ronald, Ph.D. 
digitaldog
Sr. Member
****
Offline Offline

Posts: 8027



WWW
« Reply #126 on: April 12, 2008, 06:42:06 PM »
ReplyReply

Quote
There is also  some debate at the ICC about what one should do with the demosaiced data: In which color-space one can and should encode the demosaiced scene-referred *colorimetric* data from cameras - I'm sure you're as much aware of the problems here as I am.

Not only do I not know its a problem, I don't care. I want tools that provide a desired color appearance ASAP+HQ+NBS. What happens under the hood is of little interest to me.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
eronald
Sr. Member
****
Offline Offline

Posts: 3626



WWW
« Reply #127 on: April 12, 2008, 06:46:23 PM »
ReplyReply

Quote
Not only do I not know its a problem, I don't care. I want tools that provide a desired color appearance ASAP+HQ+NBS. What happens under the hood is of little interest to me.
[a href=\"index.php?act=findpost&pid=189082\"][{POST_SNAPBACK}][/a]

Andrew, don't be disingenous. Colorimetric data needs to be saved in some space - we have nice spaces for output-referred data eg. Adobe RGB, sRGB, but nothing really future-proof and usable for the scene-referred data out of the cameras. Altough ROMM and RIMM etc were nice waystones it seems they are not quite satisfactory ...Anyway, I just hear this stuff being discussed, I don't really quite understand all of it.  I'm just a geek while all these guys are the world's geekiest color geeks

What's happening now is that the converters compute some colorimetric scene-referred data, then throw it away and retain only the output-referred rendered version.


Edmund
« Last Edit: April 12, 2008, 06:51:37 PM by eronald » Logged

Edmund Ronald, Ph.D. 
CatOne
Sr. Member
****
Offline Offline

Posts: 366


WWW
« Reply #128 on: April 12, 2008, 07:08:25 PM »
ReplyReply

Quote
Due to some info that Andrew and I are privy to, it's really, really not worth arguing about this at this stage. In the not too distant future, all of this discussion will be moot.
[a href=\"index.php?act=findpost&pid=188873\"][{POST_SNAPBACK}][/a]

Haha.  So they'll render reds as reds, or give the ability to calibrate?

Ah, we'll find out soon enough I guess  
Logged

BrianSmith
Jr. Member
**
Offline Offline

Posts: 60



WWW
« Reply #129 on: April 12, 2008, 08:18:50 PM »
ReplyReply

Holy crap!

These answers are like telling a guy who wants directions to Omaha how the internal combustion engine works...

A better name for Clarity would be Midtone Contrast.

It allows you to add contrast to the midtones without blowing out the highlights or blocking up the shadows.

Only a software engineer would call that clarity.

Yet no matter how moronic I find the name, I love the tool, so two points to the software engineers...
Logged

Peter_DL
Sr. Member
****
Offline Offline

Posts: 413


« Reply #130 on: April 13, 2008, 03:23:24 AM »
ReplyReply

Quote
As to sampling size, I would say if you flip a coin 20 times and it comes up heads 19 times, something is wrong and you do not need to extend the sample size. The thread referenced by Prof. Martinec gives good evidence of a systematic bias of a negative red hue and positive red saturation.

A typical result is shown below. The red patch (#15) is strongly shifted towards orange.
[{POST_SNAPBACK}][/a]
Of course we would first have to agree which side of the coin counts for up or down (just to follow up on this analogy).

One silent assumption with any “best-fit matrix approach” is that precisely the same definition for CIE XYZ is used a.) to shape the spectral response of the sensor according to the related color matching functions, and b.) as the master space for the camera matrix.

Can we take this for granted? Is CIE XYZ always the same?
Not necessarily. For example, the different CIE 1931 and CIE 1964 color matching functions are shown in the chromaticity.zip which can be downloaded [a href=\"http://www.efg2.com/Lab/Graphics/Colors/Chromaticity.htm]here[/url].

Just a thought.

Peter

--
« Last Edit: April 13, 2008, 03:39:20 AM by DPL » Logged
mistybreeze
Full Member
***
Offline Offline

Posts: 177


« Reply #131 on: April 13, 2008, 06:00:10 AM »
ReplyReply

Quote
Are there types of photos where use of Clarity would be of no benefit?
You have to be very careful when applying it to skin. The tiniest shadow, dimple and pore will become over-defined. Not pretty. In fact, for beauty images, I find it can be deadly, so it often stays at zero. If hard and gritty is your style, then it may have a possible use in people shots. I imagine Clarity is good on tattoos.
Logged
eronald
Sr. Member
****
Offline Offline

Posts: 3626



WWW
« Reply #132 on: April 13, 2008, 06:54:53 AM »
ReplyReply

Quote
You have to be very careful when applying it to skin. The tiniest shadow, dimple and pore will become over-defined. Not pretty. In fact, for beauty images, I find it can be deadly, so it often stays at zero. If hard and gritty is your style, then it may have a possible use in people shots. I imagine Clarity is good on tattoos.
[a href=\"index.php?act=findpost&pid=189164\"][{POST_SNAPBACK}][/a]


Maybe it's like USM ?

Edmund
Logged

Edmund Ronald, Ph.D. 
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 6767


WWW
« Reply #133 on: April 13, 2008, 08:50:13 AM »
ReplyReply

Quote
You have to be very careful when applying it to skin. The tiniest shadow, dimple and pore will become over-defined. Not pretty. In fact, for beauty images, I find it can be deadly, so it often stays at zero. If hard and gritty is your style, then it may have a possible use in people shots. I imagine Clarity is good on tattoos.
[a href=\"index.php?act=findpost&pid=189164\"][{POST_SNAPBACK}][/a]

Misty and Edmund, I don't think it's quite like USM. It is intended to (and does) accentuate very fine local differences of tonality within the mid-tones; I believe this is somewhat different than USM which accentuates any edge contrast wherever it occurs unless you deploy various masks and filters to modify it. I agree with Misty that Clarity can make skin too gritty - depending upon how strongly it is applied, because skin has varying shades and textures which would get "clarified".
Logged

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

Posts: 2714



« Reply #134 on: April 13, 2008, 09:02:44 AM »
ReplyReply

Quote
We need to do a lot more sampling of users and camera types before we go out and say for a fact there's an issue here. Lets say Adobe (Thomas) tweaks the existing profiles. Do we know for a fact that a larger audience would have other color issues? The tool exists so any user can tweak the calibration. I'd be pretty shocked if everyone needed to do this BUT Thomas. He's a pretty bright guy, knows a thing or two about image processing, Raw rendering and color management (he wrote the application that builds the Adobe ICC profiles installed, the DNG converter and originally Photoshop).

You seem pretty sure of yourself that this is a systematic issue with the calibration, with not much to back it up other than personal experience. If you have more solid numbers, we're all ears.
[{POST_SNAPBACK}][/a]

Available information is strongly suggestive that there is a systematic problem with the reds in the default ACR calibration, and no data to the contrary have been presented. However, before we jump to rash conclusions, we should look at the entire situation, not just the reds.

The following quote by DPL of Thomas Knoll is illustrative of the problem:

“Actually, to create a camera filter set that is "perfect", it is not required to exactly the match the human cone responses (or the XYZ responses). All that is required is the filter responses be some linear combination of the human cone responses. If that is the case, then a simple 3 by 3 matrix [space] can be used in software to recover the exact XYZ values.

If the filter set is not a [perfect] linear combination of the cone responses (which is the case for all current cameras), then any color calibration is going to be some kind of comprise, getting some colors correct and other colors incorrect. This is true even if you know the exact illuminant spectral curve and the exact filter spectral response curves.”


One can use a least squares method to minimize total RMS error or one can calculate the matrix coefficients to minimize error in colors that are thought to be important such as flesh tones, blue sky and foliage, which are important memory colors. One such approach is given [a href=\"http://color.psych.upenn.edu/brainard/papers/bayesColorCorrect.pdf]here.[/url]

For the Nikon D200, the overall accuracy of the default ACR rendering of the Greytag color checker is actually slightly superior to that of Nikon Capture NX even though the ACR reds have the mentioned bias.

ACR default rendering:


Nikon Capture NX (normal contrast, normal saturation setting on camera):


For your information, patches 1,2,3,4 and 15 are respectively dark skin, Caucasian skin, blue sky,  foliage, and red.

For different illuminants and different surface colors, no one set of matrix coefficients will give optimum results. Mr. Knoll's calibration likely gives the results he wants for best overall results, but it may not be optimum for reproducing the red patch on a Macbeth Color Checker.

Bill
« Last Edit: April 13, 2008, 09:04:27 AM by bjanes » Logged
Philmar
Sr. Member
****
Offline Offline

Posts: 339


WWW
« Reply #135 on: April 13, 2008, 09:33:33 AM »
ReplyReply

Excuse me people....

....shouldn't we be HAPPY that ACR renders RAW differently than DPP? Isn't it analogous to Velvia rendering differently to Kodachrome or grain film? As long as the rendering is predictable and consistent what is the fuss?

Doesn't this just mean we need to be aware of the differences and chose whatever tool we think best suits our goals?

I know different films rendered different skin tones....did photographers argue and complain endlessly about that?

All philosophical issues aside, do any of you people use the clarity slider*? Any thoughts on how it be best used? Are there some types of photos where it is best used liberally? where it should never be used?



*the subject of this thread
Logged

An office drone pension administrator by day and a photo-enthusiast by night, week-end and on vacation who carries his camera when traveling the world:
Please have a chew on my photos:
http://www.fluidr.com/photos/phil_marion/sets
madmanchan
Sr. Member
****
Offline Offline

Posts: 2100


« Reply #136 on: April 13, 2008, 09:51:03 AM »
ReplyReply

IMO, even if raw color management was a completely open standard in the early processing stages (e.g., the transform from camera coordinates to XYZ) it would still not solve the problem of dependencies on the raw converter, because of the issue Andrew points out: the color rendering. Unless the entire raw pipeline is implemented the same way across all converters (something that obviously isn't happening and won't ever happen), then raw color profiles would still be software-specific. For example, the tone curve math used by Adobe is different than Capture One's tone curve math, so even if you specified what the tone curve was in the profile you'd still have different results.

I don't know what the vendors do on the colorimetric side of things. But they certainly do things in their default color rendering which has quite little to do with colorimetry (funky hue twists, etc.). This is deliberate, by design, and is intended to provide "the Canon look" and "the Nikon look", etc.

The thing about raw color management is that it might allow one to define a standardized way of transforming camera coordinates to XYZ. But I don't see how it could usefully capture the "color rendering" aspect of color processing. And if we just omit the latter, then all we'd ever be able to do is create really consistent but visually unsatisfying output ...
Logged

mistybreeze
Full Member
***
Offline Offline

Posts: 177


« Reply #137 on: April 13, 2008, 09:53:14 AM »
ReplyReply

Quote
Any thoughts on how it be best used?
Like any creative tool, it depends on taste.

Slide it to 50 and see what it does to your image. I was quite surprised by the obvious nature of this slider: the result is apparent quickly and you get to immediately decide if it's good or bad.

I love it on landscapes. I love it on fashion. I love it on food. I love it on interiors. I love it on architecture. I love it on flowers. Did I miss anything? It's a great tool and adds just as much punch as you desire.

How far do you go? I would say that takes a bit of experimenting. You have to see the final output rendering, compare it to others, and then decide which number is best. Thank goodness for Post-It notes.  
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 8027



WWW
« Reply #138 on: April 13, 2008, 09:59:29 AM »
ReplyReply

Quote
Excuse me people....

....shouldn't we be HAPPY that ACR renders RAW differently than DPP? Isn't it analogous to Velvia rendering differently to Kodachrome or grain film? As long as the rendering is predictable and consistent what is the fuss?

Of course we should. And that said, the question doesn't address whether this is happier or not with a default setting (something that's often useless) or after moving the sliders presented to the user to produce a desired color appearance (something you couldn't do with film unless you happen to count the scanning process and the degree of alterations compared to Raw is simply tiny).

What we have is a group of people who refuse to define when they get undesirable color appearance (all the time or by default) and hence its a problem with the tool, never the user. Even if 90% of every user found a percentage of images shot that had red issues, it doesn't mean squat since:

1. They don't tell us how a profile would automatically fix this and not hose equally different types of images.

2. They don't tell us they pulled even a single slider one way or the other and made a preset.

3. They don't tell us how many other differing colors shift or don't produce a color appearance they expect in this or other converters.

In the end, its a lot of yacking from people who want to hear themselves speak or want to criticize Adobe. Is Adobe's Raw converter prefect? Nope. Have I see reds go orange? Yes and one move a slider makes that a moot point. Have I see such color issues in other converters or even in terms of scanning? Sure. Is there ever going to be even a 95% make me pretty button in a Raw converter? Nope. Rendering is subjective, that's why some of us want to work with Raw data.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
digitaldog
Sr. Member
****
Offline Offline

Posts: 8027



WWW
« Reply #139 on: April 13, 2008, 10:02:04 AM »
ReplyReply

Quote
Like any creative tool, it depends on taste

Absolutely. But some companies making software just can't win. If you put in enough granularity to produce effects but hose images in the hands of some users, said users bitch and moan. Nearly every software product that affects an image has a potential hurt me button. I'm shocked I don't hear more complain that Photoshop has a Posterize command and that using it might produce banding on your images!

The one thing software engineers can't code is taste and style. Thankfully.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Pages: « 1 ... 5 6 [7] 8 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad