Get more accurate color with camera calibration

>>the spectrum of sunlight subtly changes depending on a number of things, including the elevation, directionality of the sun (time of year in the hemispheric extremes)...

Not in terms of your camera sensor and it's interaction with a DNG profile. It is the SPD that counts: http://www.ppmag.com/reviews/200604_rodneycm.pdf
WB is very largely agnostic too, that takes place in another part of the processing path. If you view my video, you'll see that a profile made at 7000 feet can be used at sea level a few thousand miles away. My goal isn't to have you change your workflow. If you're happy, stick with it. My goal is to provide factual info about DNG profiles. And this has nothing to do with accurate (colorimetric) color, it's about producing pleasing color which is subjective. Accurate in this context is a marketing buzz word and as yet, no one has produced data that suggests this or that profile produces colorimetric accuracy.

Sorry to come to this discussion late, but in real life, my experiences match Steve's. It's true that the sun is the same color the world round, but there are clearly some independent variables out there in the real world that aren't accounted for in the ideas you talk about in your video because there are very visible differences in pictures processed with site-specific profiles vs. dual illuminants.

I find that each location--and even the same location in different seasons--has a unique effect on the color that gets reflected from objects. The good news is that profiles are easily managed and you don't have to let them clutter up your image editor.

You are correct, visually and over location, the sun differs. The spectrum doesn't such you need to build multiple DNG camera profiles. ONE for daylight, should work all day, anywhere. From Adobe engineer Eric Chan:

It's not the shooting temperature that matters, but the spectrum of the illumination. This is more complex and not easily measurable. General advice: if you tend to photograph under daylight, not worth building profile for each flavor of daylight, regardless of the color temp. There will be minor variations between ~D50 and ~D75 lighting (roughly 5000 K and 7500 K CCT) but this tends to get eliminated once WB is considered. If you photograph under unusual artifical lighting often, consider building a profile for that lighting.

The spectrum can change depending upon what is in the atmosphere filtering part of the spectrum selectively. Of course, in those situations, where the air is full of pollen (which can cause colorful halos) or dust, you may wish to use a standard daylight profile anyway precisely in order to capture the eerie effect produced by the unusual conditions.

@digidog - @VictorEngel is spot on with his comment. In real life light does not have the same spectrum just because there is a theoretical daylight spectral invariance. Environmental issues can have a huge impact - including the obvious issue of reflected color casts from your surroundings. Start going into a forest and try your luck... Now you may be happy with a "good enough" average but frankly then why bother color profiling at all? Just use a standard profile and adjust WB for visual impact - works for the 99% (and nothing wrong with that, people should just know what the trade-off is).

PS: I do not think Eric from Adobe is the best source to quote. At the end of my 5DS/R review http://www.dpreview.com/forums/post/56919326 I show the very real difference between using Adobe's standard profile and your own. Notice that I come away far better with my own profile with both highlights and shadows - at the same time(!)

>> At the end of my 5DS/R review http://www.dpreview.com/forums/post/56919326 I show the very real difference between using Adobe's standard profile and your own.

You're missing the point! There ARE differences between Adobe's profile and your own! That's why we build our own profiles. There's no need to build multiple profiles from your own sensor in differing daylight conditions (see below). ONE for daylight, one for Tungsten, one for each specific illuminant. Cleary there's a difference between an Adobe built profile using their sensor and one you build.

What Eric wrote is clear, correct and proven:

"Many flavors of natural daylight are spectrally similar (weighted differently), so this is why a daylight profile tends to work well in many flavors of daylight regardless of the actual CCT (e.g., 4700 K thru 7500 K)".

Listen to Eric, he wrote the software! Or just do the correct testing as I did in the video.

@digidog you prove nothing - and I will - later - post samples showing very clearly why what you write is either nonsense or because you choose to exclude the impact of color casts.

I proved that's not the case in my video**! Various DNG profiles built in differing locations and times of the day IN daylight produce identical results! Test it yourself.

>>PS: I do not think Eric from Adobe is the best source to quote.

You're kidding right? Do you know who Eric Chan from Adobe is?
No. I know exactly who he is. He is the person claiming that nothing is wrong with Adobe's 5DS/R profile. That's his credentials to me. Being wrong at a high technical level does not make your views more credible.

The profiles are wrong because they are inconsistent when compared to Adobe profiles for other cameras. The main issue in the result is not the cameras but the profiles.

Part of what we hope to achieve with profiles is consistency. The 5DS profiles have destroyed whatever consistency there was.

Sure, I can make my own profiles, but I didn't have to before this camera model.

BTW, taking video or pictures at various times of day does not challenge my assertion.

@digidog we simply do not have the same expectations when it comes to color profiles - so we will never agree on any of this. Enjoy your 5DS/R with the standard Adobe profiles. I'll enjoy mine with custom profiles.

Victor. Your text indicates you don't know much about profiles or their goal. A profile defines device behavior. IF the Adobe supplied profile doesn't work on your camera, there are two possibilities, one makes sense and one doesn't. What makes sense is that the one camera sample Adobe used to build their profile is quite different from yours. Again, that's WHY we have provisions to build them. The other possibility which is unproven by you, is that Adobe deliberately built a bad profile. You've missed the point again I've attempted to point out which is there is NO reason to build multiple DNG camera profiles for the same illuminant/SPD which I proved in my video by doing so! Bringing up a canned profile for a single camera that doesn't work for yours has nothing to do with the fact you do not need to build mutliple profiles for daylight, only one. Eric explained this in the quoted text and unlike you, the guy actually does understand how this all works and wrote the code to do so.

Further Victor, IF you have colorimetric proof that the canned profile Adobe built for one sensor is incorrect for that sensor, provide it. Otherwise, what you're writing is your opinion which you are entitled to but not your own made up facts!

IF a slew of Canon sensors are not consistent, which is possible, then it makes sense that a canned profile will not define those differing behaviors. IF you want to place blame, do it where it belongs, with sensors that are not consistent among a group. Of course, you have tested the profile you say is broken on how many differing Canon bodies of the the same make? One? Figures.

Digidog, your comments indicate your ignorance.

That doesn't read very well, does it? I suggest you refrain from that sort of talk and instead just discuss the issue.

Getting back to the point being discussed, I think it's very well established that the Adobe profiles for the 5DS cameras is different. They've essentially admitted that (not that it's worse, but that it's different). This is not news.

They are of course different, otherwise we wouldn't build our own. A point you continue to ignore. This has absolutely nothing to do with creating multiple DNG camera profiles for the same illuminate. Not necessary. Eric Chan who wrote the code as explained this as have I and I've shown it in an actual video.

You've got an opinion which has zero basis in both fact or colorimetric proof. We'll agree to disagree, but the more important point is, lurkers here should pretty much dismiss what you're writing as you've provided an opinion without a lick of proof. Meanwhile, I've provided both proof of concept and text from the people who write the software. I'll let those lurkers decide who they should consider having a valid point. Your points about DNG camera profiles and your inability to prove issues with the original Adobe camera sensor that built the canned profile don't have an ounce of fact or evidence.

Until you have more than your uneducated opinion, you're wasting our time.

You're right. I'm wasting my time by continuing in this thread that you persist in leaning toward ad hominem. I will agree with one thing: the readers surely can decide for themselves who has valid points. Reading list may include posts from other forums here and other sites.

>>I will agree with one thing: the readers surely can decide for themselves who has valid points.

“There are two ways to be fooled. One is to believe what isn't true; the other is to refuse to believe what is true.” 
― Søren Kierkegaard

If i can jump in:
As i can presume, digidog is probably digital dog (avatar is the same too) or Andrew Rodney, from his articles for monitor calibration etc i learned so much years ago and i was able to understand how in basic theory stuff works (manage to calibrate printer, scanner and monitor by that).
But i'm little bit puzzled now, when in comes to camera sensors (but i'm aware for camera calibration option in ps and other software for years). I do see what digidog is trying to explain about sensor and dng profiles (i't basically, how sensors react on different wavelength, not so much on shooting situations), but than i'm not seeing the final goal, calibrated colours. Ok, I understand, you manage wb and all other stuff in second phase, but then, what's the whole point? what i want from colour test card is to have colours under control or let put it this way: same results in colour fidelity, no matter which camera/sensor i'm using. Same as in monitor world..

The goal is a DNG profile optimized for YOUR sensor, not the one Adobe had when they build their "canned" DNG camera profile. But again, anyone can test this and should. AND, this is all peer reviewed though Eric Chan of Adobe in terms of the need for multiple profiles with the same illuminant. But when you read the push-back from some above of the facts, what does actual testing and info from the Adobe engineer who designed all this mean <g>?

can you explain me this: let say you get 5 guys together in same spot etc, everyone using different camera (well, basically it will be enough just using different lens manufacturer). will dng profile, made by same colour target make same results (same colour for everyone, if they all shoot same scene under same conditions)?
from video that i've seen from your link it is explained that dng profile don't manage wb an all other stuff, but what is then calibrated and where then colour target comes in?

yeah you might use many different film stocks but none of them produced truly accurate color it was about making color pleasing to the eye. I find this measuring color to the nth degree rediculous. you now have amazing control over the color look of your image your wasting it trying to measure to some perfection of realitiy

Where white balance with RAW can matter is with respect to exposure...

(1) I took a picture of a snowy mountain in the dim cloudy morning. 'Daylight' WB set. On the camera LCD, the image looked horribly blue. Looked a bit dimly exposed as well.

(2) I took a gray card reference picture with auto WB. Things look much better and no surprise, but an interesting thing happened with the histograms.

(1), the gray/white luminance histogram was pretty much in the middle. But, the blue channel was already near clipping. Wouldn't want to increase exposure.

(2), the luminance histogram was shifted a bit to the right and the image looked brighter. Blue channel had been shifted where it belongs. Didn't need to increase exposure.

WB at capture doesn't affect the RAW file, but incorrect WB at capture can affect exposure decisions during review on-site. I may make a bad decision. I can review RGB histograms on the LCD. Getting WB *mostly* right is good enough to help w/ exposure, I think.

I Googled the subject 'Does white balance affect exposure?'
and found this has been argued about for a long time now.
There's one here at DPR from 2006.
Apparently, only sunset shooters need worry (because the WB can change so quickly). Even though it's only a third of a stop difference, WB does indeed change how your camera sets its exposure.

Isnt that the purpose of wb?
Lets say you calibrate your camera using the light of the flash. Cant you then just apply white balance corrections when shooting in a different light?

So the Sigma SD1 knows about all possible light sources? I doubt it.

Are you using a web browser that recognizes monitor profiles? If not, the colors may look more vibrant than they really are.

haha, maybe we are commenting on our own monitor and its calibration, when trying to answer this question. :)

Plasnu, you must calibrate your monitor and you will see the difference. The 'after' image is definitely better than the 'before'.

That's a reasonable point. IM experience, physical Pantone calibration or comparative media require 18 month checks for accuracy. The checks are light temperature dependent.

Do you remember the discussions about how people were trying to profile their cameras because they could not get good color from ACR? PV2010 solved most of that, but still those profiles were created using only one [or maybe a couple of] camera[s] from a particular model. But there is some variation between sensors.