Luminous Landscape Forum
Raw & Post Processing, Printing => Colour Management => Topic started by: ormike on July 08, 2009, 08:50:38 pm
-
Hi Folks,
I am a printer and not a photographer by trade and I use a Canon 5D to photograph paintings under consistent studio lighting for the purpose of reproduction on a Epson 9600 or 9900.
I am searching for the best workflow to achieve ACCURATE colors as opposed to pleasing colors.
I currently use a 24 patch Macbeth color checker to build camera profiles and have used the furnished Lab target values as well as reading the Lab values of the Color Checker myself and have built numerous camera profiles using Monaco with fair but not great results.
I find I need to spend way too much time color correcting the prints in Photoshop to get the colors where they need to be which is the color of the painting.
I have used ACR with its imbedded profiles and RAW Developer with all the profiles stripped out.
I am considering going to a Hutch Color reflective target because each individual patch has been measured on each target or to the Color Checker SG target.
Actually I'm willing to try just about anything to get accurate repeatable colors.
Your comments will be very much appreciated.
-
First you will need a proper color-managed workflow from capture through output. So in addition to quality monitor and output profiles that you're probably already familiar with, you'll want a dedicated capture profile built specifically for your camera and the lights you'll be shooting under, and ideally built using the lens you'll be shooting with most of the time... This is a real icc profile you can apply to the image, not just the simple set of camera corrections you create in ACR or LR and apply during conversion. And be forewarned that unlike paper or monitor profiles, camera profiles can be a *bitch* to get just right and usually require significant time spent editing the profile for optimal results...
Cheers,
-
First you will need a proper color-managed workflow from capture through output. So in addition to quality monitor and output profiles that you're probably already familiar with, you'll want a dedicated capture profile built specifically for your camera and the lights you'll be shooting under, and ideally built using the lens you'll be shooting with most of the time... This is a real icc profile you can apply to the image, not just the simple set of camera corrections you create in ACR or LR and apply during conversion. And be forewarned that unlike paper or monitor profiles, camera profiles can be a *bitch* to get just right and usually require significant time spent editing the profile for optimal results...
Cheers,
Could you please explain how an input profile would help - or more generally, what exactly it does? I have heard people talk about input profiles (beyond the simple camera corrections) but they never made sense to me. As I understand profiles (printer and monitor), their job is to specify how the digital "colors" (numbers) in a digital image are translated to physical colors on the monitor or print. I assume that aq camera profile would come into play when the RAW image is processed, but seeing as there are no physical colors involved I don;t see what exactly it does.
THanks,
Peter
-
It does the exact same thing for your camera as it does for the monitor or paper. It tells the camera how the physical colors of the *subject* get accurately translated by the camera into those digital numbers, so that what comes out of the camera to be sent downstream is already "accurate". Now when you send that to your profiled monitor for viewing, you will be seeing what the actual subject looked like, not the camera maker's interpretation of what it should look like. Here you get to edit it to match *your* tastes whether accurate or artsy -- and the "art" is to make the digital 2-D final look more like a 3-D original (~~ an oil painting has depth to the brush strokes, a watercolor texture to the paper).
-
Have you tried Adobe's free DNG Profile Editor? The automated process still has the limitations of the 24 patch Colorchecker, but it does a good job creating a starting point, and affords the possibility of fine tuning on top of that foundation.
-
Have you tried Adobe's free DNG Profile Editor? The automated process still has the limitations of the 24 patch Colorchecker, but it does a good job creating a starting point, and affords the possibility of fine tuning on top of that foundation.
I must admit I have not done much with this editor.
I worked with it a little and then shied away because it did not allow numerical input after the fact for editing.
It seem that the DNG Editor is more for "Pleasing Color" than color accuracy but I may be wrong.
I would like to boil down my original post into the question;
Which Color Targets and Profiling Software combinations have proved to be effective in achieving Accurate Color reproduction with Digital Cameras?
Thanks,
Mike
-
Which Color Targets and Profiling Software combinations have proved to be effective in achieving Accurate Color reproduction with Digital Cameras?
http://www.basiccolor.de/english/Datenblae...t_E/input_E.htm (http://www.basiccolor.de/english/Datenblaetter_E/input_E/input_E.htm)
(homepage: http://www.basiccolor.de/english/index_E.htm (http://www.basiccolor.de/english/index_E.htm) )
-
You know, I happen to be a big fan of camera profiles, but something you said ever so casually caught my attention. " under consistent studio lighting" Can you tell me what this means? Are you using Strobes? Are the tubes all brand new. Are you using softboxes? Are these the same age?
Lighting is probably the most critical part of this equation. There is no solution better than using tungsten lights and dimmers. Strobes, no matter how good will never be a critical color match to one another. The only way to really nail this is with dimmers and a color meter. Once you match the lights you are much further down the road.
As for camera profiling, I obviously am partial to Coloreyes but two things are true no matter what you choose. A 24 patch color checker is way to small a data set and secondly unless it is a glossy target you won't have near the tonal range you need.
All that being said you will never find a solution that solves all the problems here. There will be paint colors you just can't hit and for that matter can't print. Camera profiling and critical lighting will get you dramatic time savings regardless.
-
You know, I happen to be a big fan of camera profiles, but something you said ever so casually caught my attention. " under consistent studio lighting" Can you tell me what this means? Are you using Strobes? Are the tubes all brand new. Are you using softboxes? Are these the same age?
Lighting is probably the most critical part of this equation. There is no solution better than using tungsten lights and dimmers. Strobes, no matter how good will never be a critical color match to one another. The only way to really nail this is with dimmers and a color meter. Once you match the lights you are much further down the road.
As for camera profiling, I obviously am partial to Coloreyes but two things are true no matter what you choose. A 24 patch color checker is way to small a data set and secondly unless it is a glossy target you won't have near the tonal range you need.
All that being said you will never find a solution that solves all the problems here. There will be paint colors you just can't hit and for that matter can't print. Camera profiling and critical lighting will get you dramatic time savings regardless.
Thanks Jack,
Like I said in the original post I am not a photographer by trade so the subject of lighting is as much of a challenge as learning about f-stops, ISO settings and Aperture Priority. When film transparencies became difficult to come by I chose to buy the 5D and the journey began.
When I said "under consistent studio lighting" I meant that I was shooting in the same environment. I have used tungsten with polarized filters (work-lights from Home Depot), banks of fluorescent tubes and now I am working with four Eiko EBW 500 watt daylight incandescent bulbs because I like the idea of the illuminate being as close to 5000K as possible.
Because everything you say about paint colors/pigments is true I have been shooting the Color Checker with each piece of Art and building a new profile for that shot at that time on that day.
To judge the accuracy of the profile I print the Color Checker and read the Lab numbers of the print vs. the Lab numbers of the original Color Checker and when I get some of the patches at delta E of > 20 I know something is not working right.
I agree that the 24 patches on the Color Checker is just not enough and this is why I am looking for a better target and it may well be Coloreyes. Awhile back I spoke with Derrick at Integrated Color and he stressed the importance of the lighting being even over the original art so after struggling with that for awhile I found EquaLight 2 which seems to work exceptionally well.
Again, thank you for your comments and I'll be in touch.
-
Because everything you say about paint colors/pigments is true I have been shooting the Color Checker with each piece of Art and building a new profile for that shot at that time on that day.
To judge the accuracy of the profile I print the Color Checker and read the Lab numbers of the print vs. the Lab numbers of the original Color Checker and when I get some of the patches at delta E of > 20 I know something is not working right.
That's not going to work. It doesn't tell you about the process correctly. With a product like ColorThink, you could produce the detlaE's of just the output profile, that tells you nothing about the rest of the process. You can use that product to tell you the expected and produced Lab going through the profile, that's useful for gauging the BtoA and AtoB tables in a profile. The goal for "accurate" color can't begin until you define accurate. That's not a path where the Lab values of a Macbeth in a scene are produced by measuring the Lab values of the print. Not even close. Start here:
http://www.color.org/ICC_white_paper_20_Di...ment_basics.pdf (http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf)
Got nothing to do with the number of patches on the Macbeth at this point (and yes, the DNG profile editor even with those few patches can help you produce pleasing color).
-
There is no solution better than using tungsten lights and dimmers.
If I did art repro for a living, I'd invest in a set of NorthLights: http://www.northlightproducts.com/html/copy_lights.html (http://www.northlightproducts.com/html/copy_lights.html)
However for any given copy set -- the same heads at the same distance with the same modifiers and same power settings -- the newer digital strobe packs are *very* uniform in their output, and even if you need to adjust output by one or two two stops, remain within an acceptable margin IMO. Strobe tubes will change with age however, and more so initially, so frankly it is better to start with a used set of heads.
Any way you slice it, it is a lot of work at the front end, and all you can do is get reasonably close, not perfect...
Cheers,
-
There's a lot of information here: http://www.fho-emden.de/~hoffmann/howww41a.html (http://www.fho-emden.de/~hoffmann/howww41a.html)
This pdf explains camera calibration for reproduction: http://www.fho-emden.de/~hoffmann/camcal17122006.pdf (http://www.fho-emden.de/~hoffmann/camcal17122006.pdf)
Cheers,
Kumar
-
Sorry, not even close. If we were talking about film I would agree. But having done extensive testing in this particular arena I can say without a doubt that tungstens with dimmers is the only good choice. Digital cameras have a sensitivity way beyond film and even the newest strobes are not close enough. And nylon on softboxes, no way.
As for the process of profiling you want to nail down your lighting and then make one profile. This making a profile for every shot is worthless.
If you are looking for numerical accuracy you will be spending a lot of time beating your head against concrete. Pleasing is more likely what your customer will respond too. And from my experience having built a macbeth based product and a large target product, and done extensive testing, and customer jobs on artwork and commercial color critical products, there is no comparison. Big target always wins.
-
"Tungsten with dimmers is the only good choice"
Consider the composition of the sensor, with twice as many green sensors as red or blue. So one would think that the sensor would be optimally configured for green light. Look at a light curve for a 5500 light source. The peak is in the green section. Look at the peak of a tungsten light source, red. Why do you think most digital cameras have trouble with subjects under tungsten light? So for accurate colour, you need a hotter light source. Unless you are using a colourmeter, how are you going to ensure that a tungsten light with a dimmer is operating at the proper colour temperature and how are you going to make sure it is consistent because as the voltage to the lamp changes so does the colour temperature.
You need a light source that produces a consistent colour temperature that is independent of voltage and higher than that produced by tungsten lighting for the best colour rendition.
The original poster was using tungsten halogen worklights with polarizing filters. He would be much better off getting some strobes and using them. He is going to get much less fluctuation than he will with the worklights and colour much more compatible with the sensor in the camera
-
Digital cameras have no trouble with tungsten light. Gray balance is a far better tool then you suggest. And yes you need a color meter to match tungsten lights. Strobes do not and never have produced consistent color temperature to the degree required for this kind of copywork. If I had not done a lot of this I would not bother with this any further but I've done the testing with countless photography clients and my own customers. Tungsten with dimmers is better unless power fluctuates at a given location. Camera profiling is a waste of time unless you can match the light color across the painting.
-
Jack-
I noticed on your site that you say you do not recommend using ACR or Lightroom because they do not allow the use of camera profiles. Is this still accurate for your profiling package with the current releases of those two applications, or was this for previous versions? What about Capture One? No issue there, I assume? Adobe obviously now allows the use of camera profiles, but has an odd implementation that must recognize that the profile is appropriate, and only allows for a limited number of user selected options when it comes to profiles. Do they let ColorEyes profiles (for a Phase One back, in my case) past their gatekeeper?
I use Profoto strobes with large softboxes when I can make the reflections work (shiny, highly textured canvas can be tricky, even when the boxes are flagged off) and use extremely diffused natural light otherwise. Are you saying that you have found that strobes and two layers of fresh white diffusion on softboxes still produce color variance across a scene? I normally use a light meter, but not a color meter, although I have borrowed a friend's color meter and did not read any color shifts with my setup. Obviously, some strobes are better than others. Profoto is one of the better brands, but probably not on par with Briese (or Bron?) for consistency. I observed a slight color temperature change when I reduced the power of the strobes significantly, but that was more of an FYI experiment. I don't really mess with strobe power for art reproduction. My guess is that my need for individualized color correction is far more driven by the size of the ColorChecker color sample in profiling than variance in lighting, although I wouldn't swear it isn't possible to be some of the latter.
What tungsten source do you use? HMI?
-
Jack-
I noticed on your site that you say you do not recommend using ACR or Lightroom because they do not allow the use of camera profiles.
Sure they do! Just not profiles Jack's product can build (well he could if he wanted to).
Seems an odd reason to dismiss two such products outright.
http://labs.adobe.com/wiki/index.php/DNG_Profiles (http://labs.adobe.com/wiki/index.php/DNG_Profiles)
-
Sure they do!
They do NOT work with icc profiles as input profiles and you can't process the tif in those icc profiles!
That's what "colorwave" ask for and the answer is: no.
Do they let ColorEyes profiles (for a Phase One back, in my case) past their gatekeeper?
-
They do NOT work with icc profiles as input profiles and you can't process the tif in those icc profiles!
That's what "colorwave" ask for and the answer is: no.
They allow the use of camera profiles which is what's quoted above. No, they don't allow the use if ICC camera profiles. But they of course do allow the use of custom profiles. That the product mentioned doesn't support DNG profiles in no way invalidates this approach to producing a camera profile.
I noticed on your site that you say you do not recommend using ACR or Lightroom because they do not allow the use of camera profiles
So this then begs the question, why as Colorwave asks does this web site not recommend ACR or LR? Seems pretty obvious to me.
You know, I happen to be a big fan of camera profiles...
I guess just ICC camera profiles...
-
They allow the use of camera profiles which is what's quoted above. No, they don't allow the use if ICC camera profiles. But they of course do allow the use of custom profiles.
the "profiles" you are talking about are actually just a set of internal adjustments. Begs the question if the term "profile" makes sense here.
In any case you can't process the TIFs from ACR/LR without conversion to another profile but instead just embed the camera profile (as you can do it in Capture One).
-
the "profiles" you are talking about are actually just a set of internal adjustments. Begs the question if the term "profile" makes sense here.
Maybe you can explain this further and how ICC camera profiles and DNG camera profiles are different.
In any case you can't process the TIFs from ACR/LR without conversion to another profile but instead just embed the camera profile (as you can do it in Capture One).
No, you can't embed the DNG profile as a camera profile but you can any ICC profile you have on your system. How is having the data in the camera color space beneficial? You're exporting a output referred rendered image, presumably for further editing in Photoshop, so why not be in a well behaved RGB working space?
-
The DNG Profiles FAQ contains a "question" and the answer regarding DNG camera profile vs. ICC profile; it would be interesting to see the opinions of those participants, who are dealing with these issues.
Why introduce another camera profile format instead of using ICC camera profiles?
First, ICC camera profiles used by raw converters today are designed to process output-referred (i.e., rendered) image data, not scene-referred (i.e., raw) image data. Furthermore, the sequence and placement of color transformations described in an ICC camera profile can prevent other image processing stages (such as highlight recovery algorithms) from performing optimally. Third, there is no standard that describes the input color space of the ICC camera profile color transformation (it is often, but not always, a tone-mapped set of RGB camera coordinates). Consequently, ICC camera profiles are not portable: they can only be used with the raw converter for which they were explicitly created in the first place. Using an ICC camera profile designed for one raw converter with another raw converter nearly always produces incorrect (though sometimes entertaining) results.
-
I didn't mean to stir things up with my line of questioning, but think that it has encouraged me to think about the big picture a little more.
I also thought that I might have misquoted Jack after I went back to his site again, but the exact quote is: "We do not recommend using this product with Adobe Camera Raw, LightRoom or Aperture because they do not provide for the use of camera profiles." . . . as well as . . . "Adobe has decided that camera profiles just plain don’t work and therefore they do not support them."
I think the missing modifier here is "ICC" profiles. What Adobe supports are DCP profiles. Like DNG, Adobe is trying to use their market clout to create a new standard that in practice might as well be proprietary at the moment, but is actually an open standard that they hope others will adopt. The ability to couple the profile with a Raw file for easier sharing seems to be one of the bigger selling points for Adobe on why they went their own way.
In researching DCP profiles a little more, primarily because I'd love to be able to stick with Lightroom over Capture One, an issue popped up that I hadn't heard of before. Sandy McGuffog, author of dcp Tool, says that DCP files cause a change of tint when exposure adjustments are made. I'm not sure how big an issue this is, but it certainly sounds like a concern when we are after "accurate" color, as this thread is about. His tool, which is opensource and shareware, so theoretically not particularly biased by financial motivation, makes DCP profiles "invariate", to get around this limitation. Intriguing stuff, and something I'd like to explore further, except for the fact that he says a working knowledge of XML and the DNG specifications is required, as it is a command line environment in both Macs and PCs.
How big a concern have you color experts found with the basic Adobe structural framework, if we are after "accurate" vs. "pleasing" color? There are roughly 100 to 1 discussions online discussing pleasing color over accurate, but in this case some of us are after accurate. Pleasing seems like child's play in comparison, for me, and as a Phase customer, I'm certainly not after duplicating my camera's LCD display and/or JPEG files.
For those of you interested and unfamiliar with it, Sandy has a fascinating paper online looking at how the Adobe profiles vary by twisting colors to create different looks.
You can find it here:
http://chromasoft.blogspot.com/2009/02/vis...les-part-1.html (http://chromasoft.blogspot.com/2009/02/visualizing-dng-camera-profiles-part-1.html)
I'm currently thinking an ICC camera profiling application is probably in my future, but I'd sure like to find a way to work with Lightroom if possible, over Capture One. The comparison is like OS X vs. Windows, in my book, but I'll bite the bullet if it ultimately saves me time tweaking colors one at a time.
EDIT: Here's another link to more from Sandy, which ties in with another thread here on LL regarding the profile twisting:
http://chromasoft.blogspot.com/2009/02/adobe-hue-twist.html (http://chromasoft.blogspot.com/2009/02/adobe-hue-twist.html)
One thing dcp Tool does to get around this is to remove the hue and saturation variance from value changes in the LUT. I feel a little like I'm trying to talk astrophysics with people who know it inside out, but I think I at least get this part.
-
"We do not recommend using this product with Adobe Camera Raw, LightRoom or Aperture because they do not provide for the use of camera profiles." . . . as well as . . . "Adobe has decided that camera profiles just plain don’t work and therefore they do not support them."
Which unless they update to specially state "ICC camera profiles" is nonsense. The second part (Adobe has decided that camera profiles just plain don’t work) is simply untrue. Someone is using marketing speak big time here! Adobe does of course use a camera profile as pointed out (and you mentioned above).
I think the missing modifier here is "ICC" profiles.
Part of it yes.
In researching DCP profiles a little more, primarily because I'd love to be able to stick with Lightroom over Capture One, an issue popped up that I hadn't heard of before. Sandy McGuffog, author of dcp Tool, says that DCP files cause a change of tint when exposure adjustments are made.
I'm not sure that's solely attributed to the DNG profiles. There IS and has been from day one, the effect of saturation on tonal moves in the Adobe Raw engine and there's an article here on LL by Mark Segal about this. There are reasons why this was implemented.
I'm not sure how big an issue this is, but it certainly sounds like a concern when we are after "accurate" color, as this thread is about.
How big a concern have you color experts found with the basic Adobe structural framework, if we are after "accurate" vs. "pleasing" color? There are roughly 100 to 1 discussions online discussing pleasing color over accurate, but in this case some of us are after accurate. Pleasing seems like child's play in comparison, for me, and as a Phase customer, I'm certainly not after duplicating my camera's LCD display and/or JPEG files.
Please, we need to STOP using the marketing term accurate color here. Its not accurate until you define accuracy colorimetrically and as scene referred. I posted a link to the ICC (a group that knows a few things about ICC profiles yes?) that attempts to define what accurate color really is and I can assure you, you don't want to send scene referred, colorimetrically ideal numbers to your printer. This goes back to my point about the 20 deltaE values not being a sound indicator of anything. There are all kinds of cases where correctly measured and defined "accurate" color values are NOT going to produce a match, let alone a pleasing color to the viewer. There are many cases where CIE colorimetry fails. No magic profile, input or output is change this.
-
I didn't mean to stir things up with my line of questioning, but think that it has encouraged me to think about the big picture a little more.
I also thought that I might have misquoted Jack after I went back to his site again, but the exact quote is: "We do not recommend using this product with Adobe Camera Raw, LightRoom or Aperture because they do not provide for the use of camera profiles." . . . as well as . . . "Adobe has decided that camera profiles just plain don’t work and therefore they do not support them."
The way the first sentence should read:
We do not recommend using this product with Adobe Camera Raw, LightRoom or Aperture because our products are incompatible with their camera profiles.
"Adobe has decided that camera profiles just plain don’t work and therefore they do not support them."
This is just pure, unadulterated hogwash that should just be removed from the site!
You're not to blame for stirring things up, I am. I'm just sick and tired of vendors (at least now their transparency as vendors is without question thankfully) providing technically thin, marketing thick crap to those who come to sites for information. I mean please, look at the quotes you provided and the actual reality of the situation and tell me that the BS factor is excessive.
-
How big a concern have you color experts found with the basic Adobe structural framework, if we are after "accurate" vs. "pleasing" color?
For me, it's huge and why I suggested making the profile for your camera under a specific, repeatable light source. (IME -- and admittedly I never spent tons of time trying to perfect it -- the Adobe tool does not work for accurate color: pleasing maybe, but not accurate enough for repro IMO.) The other thing you need to understand is the tool you use accuracy -- generally speaking I've found Monaco generates tighter Delta-e's on the individual colors, but Gretag generates the almost perfect grays. So the gold standard may be to build a dedicated profile with each tool, then average them... But then that's me and no doubt some of the real "experts" here will disagree
Cheers,
-
First, ICC camera profiles used by raw converters today are designed to process output-referred (i.e., rendered) image data, not scene-referred (i.e., raw) image data.
Partly right. The profiles are based on (measured) characterization data but modified (edited) to produce a certain "look". Too, at least in the case of Phase One camera profiles, they have a neutral grey axis and therefore are of course suitable for editing.
Furthermore, the sequence and placement of color transformations described in an ICC camera profile can prevent other image processing stages (such as highlight recovery algorithms) from performing optimally.
why that???
In Capture One highlight recovery works perfectly (without halos or such fancy effects).
Consequently, ICC camera profiles are not portable: they can only be used with the raw converter for which they were explicitly created in the first place. Using an ICC camera profile designed for one raw converter with another raw converter nearly always produces incorrect (though sometimes entertaining) results.
Correct.
On the other hand: processing my files in ACR without any adjustments (except pushing exposure +1 EV) as a ProPhoto TIF and then assigning my camera profile(s) in Photoshop is a much better starting point for editing as the weird colours and gradation ACR produces in default (so without extensive editing in ACR). This is far from (scene referred) acurate colours (because I have to go the circuit with ProPhoto) but it tells something about the usability of the icc camera profiles.
-
I'm not sure that's solely attributed to the DNG profiles. There IS and has been from day one, the effect of saturation on tonal moves in the Adobe Raw engine and there's an article here on LL by Mark Segal about this. There are reasons why this was implemented.
Andrew, for the benefit of those who haven't seen it and wish to, the link to my two articles on this subject is here: Curves (http://www.luminous-landscape.com/essays/curves.shtml). (One change in context is that different from what is stated there, I am now no longer a member of ACTL. )
I should add that Martin Evening took a welcome interest in that discussion and explored the subject in considerable depth, the result of which is his valuable research piece published in his Lightroom 2 book Martin Evening-Lightroom2 (http://www.amazon.com/Adobe-Photoshop-Lightroom-Book-Photographers/dp/0321555619/ref=sr_1_1?ie=UTF8&s=books&qid=1247413396&sr=1-1), pages 575 to 580 inclusive.
I think it's important in this discussion to keep separate the question of hue shift from saturation change. However, it remains true that both could reflect departure from "accurate" colour, whatever that is, because for one thing even daylight alters colour perception of light-reflecting objects in real time. I can't help but agree with most of what you have been saying (for quite some time now, not only in this thread) about the issues surrounding rendition of "accurate" color. Just thinking out loud a bit, in the grand scheme of things, what we're trying to do here is funnel into a relatively restrictive set of output conditions scenes whose colour values are valid at the moment of capture and which have tremendously greater gamut and dynamic range than anything our output devices can reproduce. Compromises need to be made all along the chain to achieve anything people would accept. Raises the question of what is acceptable? For perhaps most people, pleasing would probably be the target. For certain applied applications, such as forensics, medical (some aspects) and commercial brand recognition "accurate" may be the target.
If we're talking Adobe raw conversion software in particular, it's not as if they don't know exactly what they are doing. Tons of research and experimentation has gone into this stuff over a period of many years now and it has matured tremendously since ACR 1.0, so that while not infallible, it's probably safe to say that what they've produced IS by design. I think it's also safe to say that these tools have been designed with such tremendous flexibility that, e.g., anyone who wants to reproduce a particular hue and saturation of a particular glass of beer can do so very exactly right. For those who would find it too time-consuming to make such adjustments on each image, the software also allows the development of profiles and presets to automate an initial rendering more to ones' preference. Maybe I'm missing something, but apart from the informational question of what kind of profiles LR and ACR accept, I'm having trouble seeing what is the major problem here in using this software for achieving the colour appearance one needs.
-
For me, it's huge and why I suggested making the profile for your camera under a specific, repeatable light source. (IME -- and admittedly I never spent tons of time trying to perfect it -- the Adobe tool does not work for accurate color: pleasing maybe, but not accurate enough for repro IMO.) The other thing you need to understand is the tool you use accuracy -- generally speaking I've found Monaco generates tighter Delta-e's on the individual colors, but Gretag generates the almost perfect grays. So the gold standard may be to build a dedicated profile with each tool, then average them... But then that's me and no doubt some of the real "experts" here will disagree
Cheers,
Jack, I don't pretend to be a "real expert" on the subject of profiling, so let me ask you: you create a profile that produces what you would define as accurate colour (which is what?) under one specific lighting condition. Would that be portable (i.e. perform as well) under any other lighting conditions? When you talk about "not accurate enough for repro", what does this mean? If by "repro" you mean printing presses for example, would it be fair to say that it may not be too challenging to achieve accurate (by your definition) rendition of scene colours which are within the gamut of the output device, but what happens when they are not?
-
Jack, I don't pretend to be a "real expert" on the subject of profiling, so let me ask you: you create a profile that produces what you would define as accurate colour (which is what?) under one specific lighting condition. Would that be portable (i.e. perform as well) under any other lighting conditions?
Not really. It might perform reasonably well in brighter or dimmer values of the exact same temperature -- and why the Northlights are so good for repro -- but IME you want a dedicated profile for your standard copy sets.
When you talk about "not accurate enough for repro", what does this mean? If by "repro" you mean printing presses for example, would it be fair to say that it may not be too challenging to achieve accurate (by your definition) rendition of scene colours which are within the gamut of the output device, but what happens when they are not?
What I mean is for art reproduction -- like copying an original oil or some historical document for archival purposes. Here you are trying to capture the entire gamut of the original, and reproduce it as accurately as possible on your output media, in this case a digital print. (And we assume here the output profile is tuned for the viewing-room temp.) Most current MF cameras and scanning backs will handle enough gamut to capture anything made with historical pigments, so that's not usually a problem (though it can be for newer works). With the newer wide-gamut printers you can get very close on output, though clearly still not perfect depending on the original -- but usually close enough you need to measure the differences with your spectro because you probably won't see it with your eyes.
Cheers,
-
For me, it's huge and why I suggested making the profile for your camera under a specific, repeatable light source. (IME -- and admittedly I never spent tons of time trying to perfect it -- the Adobe tool does not work for accurate color: pleasing maybe, but not accurate enough for repro IMO.) The other thing you need to understand is the tool you use accuracy -- generally speaking I've found Monaco generates tighter Delta-e's on the individual colors, but Gretag generates the almost perfect grays. So the gold standard may be to build a dedicated profile with each tool, then average them... But then that's me and no doubt some of the real "experts" here will disagree
Couple things I don't understand:
1. What's the definition of accurate above and how do you measure/decide that ACR isn't accurate?
2. What's the delta that separates "accurate" from pleasing and how does one produce such stats?
3. I will not argue the differences between Monaco and GMB profiles (I certainly have measured data from output profiles that confirm what you say) but how does one "average" this using both, when ultimately, you have to take the average measured data and send it through one or the other package to generate a profile? I will point out too, that in the case of pleasing color output from Monaco vs. GMB, despite the delta differences which are easy to produce, often, the GMB profile produces a more pleasing print. Something useful to consider when just examining BtoA and AtoB table accuracy.
-
Please, we need to STOP using the marketing term accurate color here. Its not accurate until you define accuracy colorimetrically and as scene referred.
Andrew-
I use the word accurate in the absence of a better term for me. I'm open to a change of nomenclature, if there are better descriptions for what I want. To me, pleasing color is something with an appropriate gray balance and perceptually right looking colors that are less critical in that they are generally continuous tone, like typical photographic subjects and scenes. The scene looks right to our eye, but is not held up to the scrutiny of side by side comparison because it is impossible to compare a three dimensional scene captured elsewhere with a printed reproduction of it. Accurate color, or whatever the better way to say it might be, is color that can be looked at more critically and compared. For art reproduction, this often means solid, discrete colors that in their own way are not unlike a color chart. Artists want to be able to place a print right next to the original and see as close to the same, exact color as possible, in the same light source, on both pieces. Product photography is much the same. I realize that there are other variables in getting from an original to a final print, but I seem to have far less issues with my linearized and profiled printer than my camera. I think the greater variables come from input vs. output. I'd feel better if I could measure the RGB values of a known target and start with something in my raw converter that was extremely close, once gray balance and exposure were adjusted.
-
Andrew-
I use the word accurate in the absence of a better term for me. I'm open to a change of nomenclature, if there are better descriptions for what I want.
That's a problem because the lack of accuracy is inaccuracy and we often need some metric to define each. That's why pleasing color is more useful since all it has to do is become acceptable to you (or the client) and no measured, numeric metric is really needed. Matching color is fine too assuming you understand that two colors may appear to match which are totally different in terms of their measured color. I don't know, other than the iStar process (do a search, it was discussed here recently), we have a means of defining matching colors based on color appearance models. That is, we can have two colors that appear to match that are not accurate if we measure them and view the measured values. Which begs the question about accuracy. If they match but we have no way to define numerically using instrumentation that they match, how do we define this accuracy? When you and I both see a visual match, that an instrument tells us the values are different isn't important. We've added a true observer into the mix.
Artists want to be able to place a print right next to the original and see as close to the same, exact color as possible, in the same light source, on both pieces.
I agree that's the goal but I'd disagree that defines accurate color but not pleasing color. Maybe the word pleasing needs adjustment. But certainly, accurate is incorrect because we really don't have tools to colorimetrically define this accuracy in all cases.
-
What I mean is for art reproduction -- like copying an original oil or some historical document for archival purposes. Here you are trying to capture the entire gamut of the original, and reproduce it as accurately as possible on your output media, in this case a digital print. (And we assume here the output profile is tuned for the viewing-room temp.) Most current MF cameras and scanning backs will handle enough gamut to capture anything made with historical pigments, so that's not usually a problem (though it can be for newer works). With the newer wide-gamut printers you can get very close on output, though clearly still not perfect depending on the original -- but usually close enough you need to measure the differences with your spectro because you probably won't see it with your eyes.
Cheers,
These solutions make an attempt at accurate and pleasing ColorSage (http://betterlight.com/colorsage/index.html) and HP Artist (http://h41131.www4.hp.com/uk/en/press/hp-and-nikon-empower-new-professional-solutions.html)
Opinions?
Cheers,
Kumar
-
don't know, other than the iStar process (do a search, it was discussed here recently), we have a means of defining matching colors based on color appearance models.
Hi Andrew. I wrote the following last year on the subject, whilst discussing CIECAM02 in digital imaging. It may be of interest to you. Sorry to the rest a bit off subject.
In order to construct a chromatic adaptation model it was necessary to obtain data on the way
the visual system adapted to changes in viewing conditions. Several experiments were carried
out including those by Mori et al. from the Color Science Association of Japan, McCann et
al., and Breneman using a haploscopic matching technique; Helson et al., Lam and Rigg and
Braun and Fairchild using the memory matching technique; and Luo et al. and Kuo et al.
using the magnitude estimation method. (LUO AND CHANGJUN 2007 P263). These experiments
produced data sets called Corresponding Colours.
Corresponding Colours can be defined as a pair of colours or stimuli which are perceived to
be of the same appearance when viewed under different conditions. (FAIRCHILD 2005 P 160)
Of the methods used to determine Corresponding Colours, as mentioned above, Haploscopic
Matching involved visually isolating each eye and creating two different viewing
environments. Each eye would therefore adapt to its own environment. A test sample would
be presented on one eye and various samples presented to the other. When there was a match
the CIEXYZ value would be noted. Over time a set of Corresponding Colours would be
created. This experiment assumed that independent adaptation occurred in each eye but does
not account for the cognitive response (FAIRCHILD 2005 P160).
Magnitude Estimation involves assigning a numerical value to appearance attributes to
different stimuli. The observer is adapted to one particular viewing condition and is asked
to describe the colour in terms of Lightness, Chroma, Hue, Colourfulness or Brightness.
(FAIRCHILD 2005 P161). The experiment would be repeated using the same set of colours but
under different viewing conditions. Colours obtaining the same attribute score would be
deemed a corresponding pair. This experiment takes into account both the
sensory and cognitive adaptation mechanisms.
To produce a more natural method of matching colours, a technique called Memory Matching
was used. The observer would se a colour under one viewing condition and then try to match
the colour under the second viewing condition by using memory only. In the experiments
by Helson Judd and Warren in 1952, Munsel patches were used as the primary samples.
(FAIRCHILD 2005 P161).
From these experiments a wide range of corresponding colours data was produced. This
information was a the starting point for a chromatic adaptation model. The CIEXYZ values
of the test colour under viewing condition 1, now had corresponding CIEXYZ values of its
appearance match under viewing condition 2. The difference between XYZ1 and XYZ2 is
known as the adaptive colour shift and it is the shift that the chromatic adaptation transform
attempt to calculate. (HUNT 2004 P590)
Fairchild, M. (2005) Colour Appearance Models Reading, MA: Addison Wesley.
Hunt, R. W. G. (2004) The reproduction of Colour (6th edition), Tolworth: Fountain
Press.
Luo, M.R. and Changjun, L (2007), CIE Colour Appearance Models and Associated
Colour Spaces, Colorimetry, Understanding the CIE System, J Shanda (Ed), John
Wiley & Sons, Inc., Hoboken, New Jersey.
-
These solutions make an attempt at accurate and pleasing ColorSage (http://betterlight.com/colorsage/index.html) and HP Artist (http://h41131.www4.hp.com/uk/en/press/hp-and-nikon-empower-new-professional-solutions.html)
Opinions?
Cheers,
Kumar
Yep, Color Sage looks very impressive, but it was developed after I got out of repro so I have no direct experience with it. And I use Epson printers . But Robin likes it and IMO he is probably one of the top three or four color gurus out there -- and probably THE guy when it comes to art repro...
Thanks for the link Kumar!
,
-
Yep, Color Sage looks very impressive, but it was developed after I got out of repro so I have no direct experience with it. And I use Epson printers . But Robin likes it and IMO he is probably one of the top three or four color gurus out there -- and probably THE guy when it comes to art repro...
Agreed. In terms of very high end repro work, I can't recommend anyone higher than Robin.
-
To answer briefly some of the DCP/DNG/ICC profile questions that have come up in this thread:
Hue twists are independent of the profile format. You can have ICC camera profiles with hue twists, too (and many do). You can also have DNG camera profiles without hue twists (and many do). The term "hue twist" is probably not the most politically-adept term I could've come up with. I used the term at the time because it seemed the clearest, most direct way to describe one of the characteristics of some camera profiles; however, it draws attention (perhaps too much attention) because it sounds like something devious is going on. The truth is that many rendering systems (e.g., raw converters, or in general image processing applications doing some type of tone mapping or gamut mapping) will perform hue twists, rather often to the photographer's delight.
DNG profiles have multiple components and the color processing model described in the DNG spec allow standard raw processing steps to be inserted between those components. For example, you can perform linear light operations (e.g., if you wanted to do linear exposure adjustments, or Guillermo's Zero Noise-like compositing) in between the color matrix math and the 3D lookup table. Therefore, if you adjust exposure, and the following 3D table has colors that depend on lightness, then yes, colors may appear to shift.
(In passing, I should also mention that market clout has very little to do with Adobe's desire to use a different profile format for input. For Camera Raw and Lightroom, we do use ICC profiles where we feel they work well, technically and portably -- such as for a working space (RIMM for Camera Raw and LR) and for output (display, print, and export)).
-
Couple things I don't understand:
1. What's the definition of accurate above and how do you measure/decide that ACR isn't accurate?
2. What's the delta that separates "accurate" from pleasing and how does one produce such stats?
3. I will not argue the differences between Monaco and GMB profiles (I certainly have measured data from output profiles that confirm what you say) but how does one "average" this using both, when ultimately, you have to take the average measured data and send it through one or the other package to generate a profile? I will point out too, that in the case of pleasing color output from Monaco vs. GMB, despite the delta differences which are easy to produce, often, the GMB profile produces a more pleasing print. Something useful to consider when just examining BtoA and AtoB table accuracy.
Andrew,
From what Jack said above, it seems that the definition of "accurate" would be that in the specific case of trying to reproduce a digital image of a fine art original (e.g. an oil painting), if you were to measure the colour of a specific location on the painting with a spectro and then measure exactly the same location in the final digital print with the same instrument the readings would be the same, or so close that the eye would not notice the difference if the two were compared side by side (complicating operational factor: can the spectro actually measure accurately the colour of light reflected by these two very different kinds of materials?). In Jack's example, the colours of the original painting are not out of the gamut of say an Epson 7900, so gamut compression math would not be called upon and therefore nothing here to create the necessary compromises between "accurate" and "pleasing". But if there were out of gamut colours then either Perceptual or RelCol would produce (different kinds of) "inaccuracy" which by the logic of the process seems inescapable regardless of what profiling one does - so better for purposes of this discussion stick with a constraint of "everything is within printer gamut".
Now, if I were photographing paintings (and I used to do this with 4*5 inch Ektachrome sheet film I processed myself (for "accuracy" - hah) decades before photographer-controlled colour management was a gleam in our eyes) today, I would do this digitally with "daylight" floods" but also do test shots with a GMCC and use the second to lightest grey patch for neutralizing (if any) colour cast using the eyedropper tool in LR 2.4. We don't have control over what the camera's firmware does with the raw image data before it is exported to the raw converter, so we're talking about what happens in the raw converter. When we do the grey-balancing, we (speaking for myself and I suspect a great many others) are making the presumption that the math in the raw converter corrects every other colour commensuately such that all those colours become "accurate". So that's the question - if by design it's not happening in LR/ACR's default algorithms - is it possible to achieve such accuracy (here asking Eric or others who also actually work the math of this stuff) using the profile modifcation steps which Eric mentions above? And if so, for the benefit of those who need this flavour of "repro accuracy", could we be pointed to some specific instructions for actually implementing it?
Mark
-
From what Jack said above, it seems that the definition of "accurate" would be that in the specific case of trying to reproduce a digital image of a fine art original (e.g. an oil painting), if you were to measure the colour of a specific location on the painting with a spectro and then measure exactly the same location in the final digital print with the same instrument the readings would be the same...
You sure about this?
-
You sure about this?
with the quote completed ...
or so close that the eye would not notice the difference if the two were compared side by side (complicating operational factor: can the spectro actually measure accurately the colour of light reflected by these two very different kinds of materials?)
... I'd agree with Mark here.
Especially regarding "or so close that the eye would not notice the difference".
Maybe this could be a good intended use for the Colormunki with its UV cut filter (am not sure)...??
-
with the quote completed ...
... I'd agree with Mark here.
Especially regarding "or so close that the eye would not notice the difference".
Maybe this could be a good intended use for the Colormunki with its UV cut filter (am not sure)...??
There are a few potential issues. First is the illuminant used in the Spectrophotometer and the possible disconnect between the illuminate of the viewing conditions presented to the observer. Next there's the possibility of metameric failure or simultaneous contrast, a pretty common phenomena that cause the same color under the same illuminant to appear differently depending on the background color against which it is viewed. And again, we're assuming that taking a spot color reading of a complex image, and an associated spot reading on output, using CIE Lab colorimetry is going to tell us there's a match (rather than just looking at the two).
Bruce Fraser as usual, summed it up well and in a fashion that's easy to understand:
So while Lab is useful for predicting the degree to which two sets of
tristimulus values will match under very precisely defined conditions that
never occur in natural images, it is not anywhere close to being an adequate
model of human color perception. It works reasonably well as a reference
space for colorimetrically defining device spaces, but as a space for image
editing, it has some important shortcomings.
The viewing conditions for which Lab was designed will never exist in a
natural image. I repeat, Lab was designed to predict the degree of matching
between two solid color patches of a certain size on a neutral background.
Viewing conditions aren't just a matter of the illuminant, but also opf the
surround, and the spatial organization of the target colors. It's an
exaggeration to say that Lab describes colors. What Lab describes is
tristimulus values that, under the specified viewing conditions, will be
seen by the "standard observer" as colors. But when these tristimulus
values are located in a natural image, the perceived color is subject to
all the perceptual phenomena I've been mentioning, and the perceived color
may be quite different from what Lab predicts. If the color is
colorimetrically correct, but looks wrong, the image doesn't work.
-
In general, creating these so-called "accurate" profiles is certainly possible, within limits. I say so, without defining the term accuracy, because for any reasonable definition of accurate, one can usually optimize the profile to meet the definition. However, the caveat is that the technical information required to build such a profile is usually not available, at least not easily, to end users. Ideally you would have spectral information regarding the scene characteristics (i.e., a set of spectral radiance samples of the scene you wish to photograph and reproduce) and spectral information about the optical system and sensor. The ColorSage solution referenced above is an example that has this, in a specific context (specific camera, specific scene); their overall methodology would generalize well to other contexts.
-
Eric,
Thanks for clarifying this. What you saying seems to suggest that at this time it isn't possible to have a generic profile for use in a raw converter which will deliver "accurate" colour for just about any scene. Please correct me if I'm reading wrongly.
Andrew,
No, of course I'm not sure - that's why I used qualified language, and yes, your quote from Bruce Fraser is indeed telling.
Mark
-
There are a few potential issues. First is the illuminant used in the Spectrophotometer and the possible disconnect between the illuminate of the viewing conditions presented to the observer. Next there's the possibility of metameric failure or simultaneous contrast, a pretty common phenomena that cause the same color under the same illuminant to appear differently depending on the background color against which it is viewed. And again, we're assuming that taking a spot color reading of a complex image, and an associated spot reading on output, using CIE Lab colorimetry is going to tell us there's a match (rather than just looking at the two).
Bruce Fraser as usual, summed it up well and in a fashion that's easy to understand:
Thanks, Andrew.
I think that's basically clear.
But beside the limitations of colometric measurement there is still the criterion of visual perception. If you compare the reference artwork and the print you'd like them to match. Here it's not so essential that two colours match colormetric. The impression of "matching" is enough here.
Now the color measurement might be limited but at the same time it helps a lot in getting close to what is required. Sometimes very close (e.g. a Monitor and a print viewed under appropriate conditions can match very well though the display has backlight and the print has incident light).
Simplified: the closer you get with colormetric measurement the better. The rest is visual tweaking of colours and gradation in Photoshop.
And here I think icc profiles produced with e.g. "BasICColor Input" (trades under the name Color Eyes in the US, AFAIK) or similar products are a HUGE help to get close. If you say there is no matching based on colormetric measurement and therefore you don't need suitable input profiles I think this is a bit of a stretch (at least that's a bit how I am reading your comments - correct me if I'm wrong).
It is a fact that once you open an (Phase One DB) image in Capture One that it actually "reminds" you of the scene (the tradition of film conventions affect our perception as well, of course); maybe not "matching" by default, but not too far away. I once created a profile with "BC Input" and it was shocking how close the profile was to the reference. Problem here was: it was also shocking how unusable it was under different lighting conditions :-) This is why I stopped learning to create camera profiles and just use the Phase presets (Tungsten, Daylight, Outdoor Daylight, Flash ...) for my cameras. But for repro work I certainly would try it this way.
-
Simplified: the closer you get with colormetric measurement the better.
Again, not necessarily, maybe, sometimes.
The rest is visual tweaking of colours and gradation in Photoshop.
That's probably going to be necessary in many, many cases.
And here I think icc profiles produced with e.g. "BasICColor Input" (trades under the name Color Eyes in the US, AFAIK) or similar products are a HUGE help to get close.
Maybe. I think the idea that the input profile alone deserves sole credit here might be part science and part marketing. Clearly these profiles play an important role. But they can (and in some cases have) been over sold.
If you say there is no matching based on colormetric measurement and therefore you don't need suitable input profiles I think this is a bit of a stretch (at least that's a bit how I am reading your comments - correct me if I'm wrong).
That's not what I'm saying. I'm saying unless we know the entire processing path to the degree someone like Eric does in his product, we're all unqualified to say to what degree "accurate" color is produced solely by some input profile.
It is a fact that once you open an image in Capture One that it actually "reminds" you of the scene (the tradition of film conventions affect our perception as well, of course); maybe not "matching" by default, but not too far away.
Scene referred? Reminds you based on what? Did it record the illuminate of the scene and does it hold the spectral sensitivity of your chip in order to produce this report, or, it gives you the impression it knows about the scene? I don't know, I don't use that product but I'd be hard pressed to believe it has the necessary information that Eric and Robin are talking about ala ColorSage.
Its useful to at least attempt to separate the marketing hype or speak from the actual underlying functionality here. We've as yet not even defined the metric of "accurate" color here. Even if we use the term matching color, we need to agree that what matches for one observer may not for another, or in another environment. This is pretty complex stuff. But someone selling you a product can easily gloss over all this and just tell you "build a profile (ICC or otherwise) and you'll have accurate color. That raises the hair on the back of my neck every time I hear this.
-
In general, creating these so-called "accurate" profiles is certainly possible, within limits. I say so, without defining the term accuracy, because for any reasonable definition of accurate, one can usually optimize the profile to meet the definition. However, the caveat is that the technical information required to build such a profile is usually not available, at least not easily, to end users. Ideally you would have spectral information regarding the scene characteristics (i.e., a set of spectral radiance samples of the scene you wish to photograph and reproduce) and spectral information about the optical system and sensor. The ColorSage solution referenced above is an example that has this, in a specific context (specific camera, specific scene); their overall methodology would generalize well to other contexts.
Are the spectral information of camera sensors known to RAW converter developers? Or maybe the influence of various lenses to spectral sesitivity function is too significant?
Is it possible and effective to utilize spectral measurements of ambient light to create camera profiles for popular DSLR cameras?
Could the spectrophotometer become photographers best friend in the future? Could it solve the AWB issue somehow?
-
Is it possible to utilize spectral measurements of ambient light to create camera profiles for popular DSLR cameras?
Could the spectrophotometer become photographers best friend in the future? Could it somehow solve the AWB issue?
Yes and yes but it could be expensive. Eric Walowit who's on the ICC Digital Camera group has for years proposed the idea that our cameras would record the spectral properties of the scene, embed this as EXIF data with the Raw and, along with the important spectral sensitivity of the camera, build an on-the-fly profile for each image. It makes a lot of sense when technology affordable and dismisses Jack's idea about "one profile for all scenes" but that's admittedly based on the current and not very robust solutions offered today.
-
Reminds you based on what?
reminds me of the scene - the way I perceived it. I "accept" the colours and gradation as a photograph that represents something close to what I've seen.
I basically agree to the most of what you say. But whilst others define "accurate" we keep on photographing and I prefer input profiles that produce (howsoever) "useful" colours to start with. And I have yet to see that I open a RAW file in ACR/LR and it's by default close to the scene in some way (certainly depends on the camera... I don't know). That's what I don't get and my read is that at Adobe they share your standpoint that unless we can't do it really right, we cancel it all the way. Of course in C1 colours does not match "accurate" by default (how could they)... but you won't wonder about the colours you see (mostly) once you open a RAW file and set WB.
Its useful to at least attempt to separate the marketing hype or speak from the actual underlying functionality here. We've as yet not even defined the metric of "accurate" color here. Even if we use the term matching color, we need to agree that what matches for one observer may not for another, or in another environment. This is pretty complex stuff. But someone selling you a product can easily gloss over all this and just tell you "build a profile (ICC or otherwise) and you'll have accurate color. That raises the hair on the back of my neck every time I hear this.
I agree.
-
It makes a lot of sense when technology affordable and dismisses Jack's idea about "one profile for all scenes" but that's admittedly based on the current and not very robust solutions offered today.
Uh excuse me, but where did I ever indicate "one profile for all scenes" ??? What I said above is a dedicated profile for a given set of lighting...
Here is what I said above:
Not really. It might perform reasonably well in brighter or dimmer values of the exact same temperature -- and why the Northlights are so good for repro -- but IME you want a dedicated profile for your standard copy sets.
Let me repeat that last part in BOLD: but IME you want a dedicated profile for your standard copy sets.
Cheers,
-
reminds me of the scene - the way I perceived it. I "accept" the colours and gradation as a photograph that represents something close to what I've seen.
Yes, that's nice but playing devils advocate, its again nearly impossible to say this is true or not. Its like those that want the print to match the display but don't have proper viewing conditions for viewing the print. They take the print to a number of locations away from display and while that lighting might be ideal, with out the display and print side by side, you're reduced to "memory" colors of what does or doesn't match. In the above discussion, we're at least talking about viewing a reproduction and the original under the same illuminant and asking if they match. Imagine if your client not took the fine art print home and said "it doesn't match" without having the original next to it (or a more common problem that might be impossible to fix, client finds that print and painting under different illuminate now do not match when they did in your studio).
Getting back to accurate, its by definition in a measurable way, the scene referred capture and by the time the Raw converter renders an output referred preview on screen, its just impossible to say its accurate or a match. In the ICC article link above, the reasons are explained.
And I have yet to see that I open a RAW file in ACR/LR and it's by default close to the scene in some way (certainly depends on the camera... I don't know). That's what I don't get and my read is that at Adobe they share your standpoint that unless we can't do it really right, we cancel it all the way. Of course in C1 colours does not match "accurate" by default (how could they)... but you won't wonder about the colours you see (mostly) once you open a RAW file and set WB.
Now we have to investigate if this problem is solvable using either or both custom or tweaked DNG profiles or preset rendering adjustments. Its possible this isn't possible. Its possible C1 produces a more preferable match by default (there have been lots of users who complain about the default ACR rendering as such, the defaults have been tweaked over the years. You can't please all the people all the time). I'm not disagreeing with your findings here. I prefer the default rendering of Raw Developer and taking color and tone out of the mix, there's just some Raw processing that I find preferable that no amount of profile or rendering tweaking in ACR will match.
-
Uh excuse me, but where did I ever indicate "one profile for all scenes" ??? What I said above is a dedicated profile for a given set of lighting...
Other Jack. See above.
-
Yes and yes but it could be expensive. Eric Walowit who's on the ICC Digital Camera group has for years proposed the idea that our cameras would record the spectral properties of the scene, embed this as EXIF data with the Raw and, along with the important spectral sensitivity of the camera, build an on-the-fly profile for each image. It makes a lot of sense when technology affordable and dismisses Jack's idea about "one profile for all scenes" but that's admittedly based on the current and not very robust solutions offered today.
Thanks Andrew - I know many wedding photographers, that would definitely like Eric's idea - even if they had to pay a fortune and mount a ColorMunki (or even i1pro) on top of their cameras...
Now I'll patiently wait for i1pro support in DNG Profile Editor
-
Other Jack. See above.
Duh. Sorry
-
Duh. Sorry
I should have been more clear. No worries.
-
Yes, that's nice but playing devils advocate
but at the current state of colormanagement it's sometimes all we can do :-) A friend of mine is a prepress guy and they do a lot of shots for sports products... all synthetics of course (okay, nasty). It's impossible to measure the (product-) colours under D50 or even under D65. No way. He takes the products, goes out of the office and eye up the stuff under daylight in different conditions. Then he tweakes the colours from memory.
-
And using a printer profile built on one rip won't work on another so I guess I don't see the point.
-
Yes, Mark, you understood me correctly. Camera profiles are generally built with a set of assumptions in mind. Being pre-baked, without detailed knowledge of the photographer and his/her preferences, the scene content to be photographed, or the illumination to be used, the assumptions can only go so far. The strength of the ColorSage solution is that many of the variables are removed: information about the surface characteristics, the lighting used to capture the image, the lighting used to reproduce the image, and the characteristics of the optical system are all provided. For most general purpose photography, such detailed information is not always available, though the steps proposed by Eric Walowit (outlined by Andrew) are in the right direction.
This is one reason why most cameras provide multiple rendering presets selectable from the menu options, with names like Standard, Neutral, Natural, Vivid, Portrait, Landscape, Monochrome, Faithful, etc. Even assuming detailed scene characteristics are available, the camera designers know that photographers will simply have different individual preferences and cannot predict for a given exposure how the user will want to render that image.
-
Thanks for the confirmation Eric.
So many variables - like peeling an artichoke - you can keep going till you get down to the core, by which time the problem is reduced to a very small manageable set of conditions which can be targeted and programmed for predictable, controlled outcomes. For a great many uses of photography that simply isn't necessary, and for those needing it - well they can peel the artichoke and craft their presets accordingly. So in this context, one of the most important characteristics of raw conversion software becomes its ability to respond flexibly to different needs, and from what I observe, that is the direction in which LR/ACR have been moving.
-
So in this context, one of the most important characteristics of raw conversion software becomes its ability to respond flexibly to different needs, and from what I observe, that is the direction in which LR/ACR have been moving.
...but some kind of illuminant's spectral data utilization in DNG Profile Editor would also be nice...
-
Hi Guys
Ive been following this thread with interest. It looks like there is some grey ares in defining 'accurate' and pleasing etc.
Ive been working in colour appearance (based on our perception) as opposed to "accurate colormetric colour" (if that the right term!)
Yes and yes but it could be expensive. Eric Walowit who's on the ICC Digital Camera group has for years proposed the idea that our cameras would record the spectral properties of the scene, embed this as EXIF data with the Raw and, along with the important spectral sensitivity of the camera, build an on-the-fly profile for each image. It makes a lot of sense when technology affordable and dismisses Jack's idea about "one profile for all scenes" but that's admittedly based on the current and not very robust solutions offered today.
I finished some interesting work last year based on the above ideas by Eric Walowit who gave us a presentation in London in 2007t. I introduced a colour appearance model (CIECAM02) into the raw digital pipeline. The camera was spectrally characterised using a double monochromator and a matrix produced using Erics methods.
The colour appearance model uses CIE colorimerty to deternmine parameters etc but then mathematicly models human perception.
Camera Nikon D70s was used as Eric had spectral data for that camera type.
To cut a long story short the preocess involved obtaining RAW file using DCraw, applying an optimised camera matrix (using Erics methods) to obtain scene referred colorimetry and the applying CIECAM02 and then rendering to sRGB (easiest option for evaluation purposes).
I obtained excellent results and compared with camera JPEG there was a real difference. The scene was more as I remembered it. I tried to replicate the results using ACR and NIkon Raw developer. Got nearly there but could mot match CIECAM02 for improvement in colour.
Ill post some examples if anyone is interested.
Colour appearance processing is the way forward but there are a few issues still needing ironed out first, but first results are promising.
I hope to publish a paper later this year at Digital Futures 2009 in London
digital futures (http://www.rps-isg.org/DF2009CallPapers.php)
Iain
-
Hi Guys
Ill post some examples if anyone is interested.
Colour appearance processing is the way forward but there are a few issues still needing ironed out first, but first results are promising.
I hope to publish a paper later this year at Digital Futures 2009 in London
digital futures (http://www.rps-isg.org/DF2009CallPapers.php)
Iain
Please do and let us know when the paper is available.
-
The colour appearance model uses CIE colorimerty to deternmine parameters etc but then mathematicly models human perception.
Camera Nikon D70s was used as Eric had spectral data for that camera type.
To cut a long story short the preocess involved obtaining RAW file using DCraw, applying an optimised camera matrix (using Erics methods) to obtain scene referred colorimetry and the applying CIECAM02 and then rendering to sRGB (easiest option for evaluation purposes).
I obtained excellent results and compared with camera JPEG there was a real difference. The scene was more as I remembered it. I tried to replicate the results using ACR and NIkon Raw developer. Got nearly there but could mot match CIECAM02 for improvement in colour
Fascinating!
Could you foresee, if the introduction of CIECAM02 in ACR/LR or Nikon Capture alone would help to achive comparable results, or would it also be absolutely necessary to optimise color matrices using Erics methods?
-
Fascinating!
Could you foresee, if the introduction of CIECAM02 in ACR/LR or Nikon Capture alone would help to achive comparable results, or would it also be absolutely necessary to optimise color matrices using Erics methods?
I think that to use CIECAM02 effectively requires accurate spectral characterization of the camera (which should include lens and any filter. I performed characterization using the lens supplied with the Nikon with a skylight filter in place. I dare say changing components will have some effect but how much I dont know).
Methods of calculating a matrix are published in ISO 17321-1:2006 . I dare say that other ways of optimising the matrix can be evolved and incorporated into a RAW workflow. The target based characterization used by ACR at the moment might not be sufficient to produce the required results. I also did a target based characterisation and obtained a matrix but found the approach not as flexible as the spectral based method. So there may be issues there.
Camera manufactures who have accurate spectral characterisations of their products would be in a very good position to develop appearance processing. As for third party developers, this would require a bit more work, but not impossible!
But having said that I cant see it being too long before the idea of CIECAM02 processing catches on. (its in Windows Vista) There is still a bit of work to do as i said to get a robust camera to print workflow. Main stumbling blocks is defining the viewing conditions of the scene and those of the output. CIECAM02 requires these to be defined accurately for optimal results. I used the output conditions defined for a sRGB display (ISO 3664:2009).
I think that there are ways of incorporating the necessary information into an ICC profile, but then arises the need for smart CMMs (another issue altogether).
Using CIECAM02 for general outdoor photography relies heavily on the white point estimation. Using CIECAM02 in a controlled studio environment where the white point can be measured accurately would be ideal. And if the output viewing conditions were controlled also (in the tube system for example, paper media and emissive displays under uniform lighting spec) a well controlled appearance chain could be set up.
Studio photography would benefit greatly from the use of appearance models as would consumer level cameras where sRGB JPEG is the output referred medium.
-
I think that to use CIECAM02 effectively requires accurate spectral characterization of the camera (which should include lens and any filter. I performed characterization using the lens supplied with the Nikon with a skylight filter in place. I dare say changing components will have some effect but how much I dont know).
Methods of calculating a matrix are published in ISO 17321-1:2006 . I dare say that other ways of optimising the matrix can be evolved and incorporated into a RAW workflow. The target based characterization used by ACR at the moment might not be sufficient to produce the required results. I also did a target based characterisation and obtained a matrix but found the approach not as flexible as the spectral based method. So there may be issues there.
Camera manufactures who have accurate spectral characterisations of their products would be in a very good position to develop appearance processing. As for third party developers, this would require a bit more work, but not impossible!
But having said that I cant see it being too long before the idea of CIECAM02 processing catches on. (its in Windows Vista) There is still a bit of work to do as i said to get a robust camera to print workflow. Main stumbling blocks is defining the viewing conditions of the scene and those of the output. CIECAM02 requires these to be defined accurately for optimal results. I used the output conditions defined for a sRGB display (ISO 3664:2009).
I think that there are ways of incorporating the necessary information into an ICC profile, but then arises the need for smart CMMs (another issue altogether).
Using CIECAM02 for general outdoor photography relies heavily on the white point estimation. Using CIECAM02 in a controlled studio environment where the white point can be measured accurately would be ideal. And if the output viewing conditions were controlled also (in the tube system for example, paper media and emissive displays under uniform lighting spec) a well controlled appearance chain could be set up.
Studio photography would benefit greatly from the use of appearance models as would consumer level cameras where sRGB JPEG is the output referred medium.
Thanks a lot for interesting answer, I hope I'll see CIECAM02 processing in Adobe apps soon.
As for spectral characterisation - I wonder if it would be feasible, that some manufactrer would offer inexpensive monochromators for personal use - or maybe it would be too complicated/expansive for mass production?
-
Thanks a lot for interesting answer, I hope I'll see CIECAM02 processing in Adobe apps soon.
As for spectral characterisation - I wonder if it would be feasible, that some manufactrer would offer inexpensive monochromators for personal use - or maybe it would be too complicated/expansive for mass production?
My limited understanding from Eric and others is, the spectral data is key here, the CIECAM02 less so (but very, very useful as its based on appearance modeling).
-
Here is one I did earlier
Images have been taken on the NIKON D70s set to automatic mode – ‘point and shoot’
– with no exposure compensation used in any of the scenes.
The file has three images contained in it.
Ist layer is Nikon JPEG as shot.
2nd is my attempt to recreate scene using ACR from .NEF file
3rd is CIECAM02 processed from .NEF file, with this layer is a levels adjustment. The processed files were coming out of CIECAM02 dark. I did not introduce exposure rendering into the algorithm before converting to sRGB, opting to adjust in photoshop. (dont ask, had a few issues in this area but more interested in what CIECAM02 was doing colour wise). There is no other post processing involved.
So adjust levels to suit - I have my levels marked at the side - it would be interesting if I can get what your optimum level adjustments are.
I viewed this in dim environment on a sRGB monitor with a white point of D65, with the imaged toggeled to a black background in photoshop and all but the levels palette showing at the side.
Please give yourself time to adapt to the dim environment then toggle between layers to see the effect.
Woods near Inverness Scotland.
CAMERA SETTINGS • ISO 1600 • f 22 • 1/30s • 18mm WB 4650
CIECAM02 • La 500 • Yb 20 • ‘avg’ • Matrix D55 La 780
This is a typical Highland woodland scene. It was taken on a very bright day causing the
leaves to be backlit. There is not much white in the scene and the luminance levels in the
actual wood have fallen to about 500cd/m2. CIECAM02 has again has opened up the
shadows and has definitely increased overall colourfulness. This is especially noticeable in the
small areas of sky.
See what you think. Ill dig out some more.
Couldnt get pic to attach, dont think it likes tiffs so ive put int on my website
sample ciecam02 (http://www.hiweb.co.uk/htmlfiles/cam02.html)
-
CIECAM02 is often used internally when building profiles (i.e., as the space in which residual errors are minimized). See ProfileMaker, for example.
-
See what you think. Ill dig out some more.
Couldnt get pic to attach, dont think it likes tiffs so ive put int on my website
sample ciecam02 (http://www.hiweb.co.uk/htmlfiles/cam02.html)
It looks very promising, indeed. The colors are vivid, and hues are very well separated. I like the way the lights were rendered - the moss and leaves in sun stains kept intensive colors, and well defined details. I'd be glad to see some other examples.
CIECAM02 is often used internally when building profiles (i.e., as the space in which residual errors are minimized). See ProfileMaker, for example.
I've also noticed mysterious "CIECAM02" here and there, for example in basICColor display, and Argyll CMS.
A while ago I also found an interesting plug-in for Photoshop (freeware, Windows only):
http://cliff.rames.googlepages.com/ciecam02plugin (http://cliff.rames.googlepages.com/ciecam02plugin)
-
hi Czornyj
it was a first pass thru as to speak.I would say I could perhaps improve, but the exercise was to see what CIECAM02 could do rather than get a first "bang on" result. it takes a bit of getting used to the nuances of the input parameters and their subtle influences on the final result.
not bad for a first attempt!
I chose this scene as i wanted to see how the model could cope with a difficult lighting situation, especially with the shadows.
-
Iain,
I'm curious about the details of your method. You say that "applying an optimised camera matrix (using Erics methods) to obtain scene referred colorimetry and the applying CIECAM02 and then rendering to sRGB (easiest option for evaluation purposes)". It seems to me that that is in effect no different to, as Eric Chan says above, using CIECAM02 to generate a profile, then using that profile. Or am I missing something?
Sandy
-
Iain,
I'm curious about the details of your method. You say that "applying an optimised camera matrix (using Erics methods) to obtain scene referred colorimetry and the applying CIECAM02 and then rendering to sRGB (easiest option for evaluation purposes)". It seems to me that that is in effect no different to, as Eric Chan says above, using CIECAM02 to generate a profile, then using that profile. Or am I missing something?
Sandy
I think what Eric Chan means is that the gamut mapping / rendering is done in CIECAM02 appearance coordinates of JCH (lightness, Chroma and Hue) as this is a perceptually better space than CIELAB.
What can happen is that XYZ is converted to JCH and back to XYZ with the same input and output CIECAM02 parameters, so in effect no actual appearance processing is taking place, only a conversion to another colour space.
And yes the scene referred colorimetry and the viewing conditions can be saved into a profile for later processing in situations where the output viewing conditions and device are not known. There are two tags in the ICC v4 profile spec (page96) for viewing conditions:
viewingCondDescTag - viewing condition description
viewingConditionsTag - Viewing condition parameters.
So the framework is there, just needs to be put into use.
Ps added another sample pic rainbow over Crystal Palace London (http://www.hiweb.co.uk/htmlfiles/cam02.html)
-
I've not really been following all of this topic as it's been a bit over my head, but looking at your images there seems to be a few major downsides to the CIECAM02 rendering.
There seems to be significantly more chroma noise, an artefact that looks like really bad chromatic aberration around fine details, and looking at the histogram is a bit concerning. I'm also about curious—are the “ACR from NEF” layers using a custom profile, or just one of the standard ones? (I found that none of the Adobe ones were any good with my camera)
-
I've not really been following all of this topic as it's been a bit over my head, but looking at your images there seems to be a few major downsides to the CIECAM02 rendering.
There seems to be significantly more chroma noise, an artefact that looks like really bad chromatic aberration around fine details, and looking at the histogram is a bit concerning. I'm also about curious—are the “ACR from NEF” layers using a custom profile, or just one of the standard ones? (I found that none of the Adobe ones were any good with my camera)
You should take into account, that - like usual in such cases - Iain used the open source DCRAW converter as a base to further experiments, so chroma noise and chromatic aberration have nothing to do with CIECAM02 (it's only a color model), but it's simply a matter of demosaicing algorithm of DCRAW (that doesn't handle noise and aberration as good as commercial Nikon or Adobe products).
Apart from noise etc., the samples look really interesting - the colors are clean, vivid and well separated, even in the shadows and highlights areas. To my eye these pictures look atractive, but also suprisingly realistic.
-
There seems to be significantly more chroma noise, an artefact that looks like really bad chromatic aberration around fine details, and looking at the histogram is a bit concerning. I'm also about curious—are the “ACR from NEF” layers using a custom profile, or just one of the standard ones? (I found that none of the Adobe ones were any good with my camera)
Hi Andrew. Any noise in the original RAW will be accentuated by CIECAM02. Noise is not a CIECAM02 artifact. I used DCRAW as it exports demosaiced RAW files, my starting point. I cant get this from NIkon or ACR.
The ACR from NEF layer was my attempt to replicate the scene using ACR. All adjustments were were manual, no profiles.