Luminous Landscape Forum
Raw & Post Processing, Printing => Colour Management => Topic started by: torger on October 03, 2013, 03:09:58 am
-
Help me out with some color management basics;
As far as I understand most systems are calibrated for some standard light, say daylight or flash, 5500K or so. With that kind of illuminant I guess it's possible to get accurate colors out of most cameras with appropriate profiles
However, what happens if you shoot in the dusk or dawn, or like me in the winter above the arctic circle, the color temperature of the light is then far from the standard 5500 or 6500K. I have a hunch that this affects how the eye/brain experience colors so you would need some other color model for these temperatures in order to make a calibration to accurately support that. Is there such a thing, or is accurate color management locked to illuminants close to 5500K?
Why would I want to have accurate color in extreme color temperatures you ask? If I could I would, I love to have "neutral" as a starting point when I process my pictures, but I have noted that for these light conditions it's not easy. I'd just like to know if it's possible with current technology and color models.
-
Hi,
I don't think accurate color is possible in any light...
In general, I think white balance is most important. Just getting it right helps a lot, but a surface in shadow or sunlight will have a different white balance, in many pictures you have both.
Sometimes a color profile for specific lighting can be helpful, but we still have metameric issues. To produce accurate colora you would need a well defined target with the same spectral response for each color as in your subject, spectral data for that target and a program generating accurate profiles from arbitrary data.
Best regards
Erik
Help me out with some color management basics;
As far as I understand most systems are calibrated for some standard light, say daylight or flash, 5500K or so. With that kind of illuminant I guess it's possible to get accurate colors out of most cameras with appropriate profiles
However, what happens if you shoot in the dusk or dawn, or like me in the winter above the arctic circle, the color temperature of the light is then far from the standard 5500 or 6500K. I have a hunch that this affects how the eye/brain experience colors so you would need some other color model for these temperatures in order to make a calibration to accurately support that. Is there such a thing, or is accurate color management locked to illuminants close to 5500K?
Why would I want to have accurate color in extreme color temperatures you ask? If I could I would, I love to have "neutral" as a starting point when I process my pictures, but I have noted that for these light conditions it's not easy. I'd just like to know if it's possible with current technology and color models.
-
I agree with Erik's first sentence. I think too many photographers waste their time in pursuit of "accurate". I read a few days ago in a thread in the site that the colour cards on sale aren't all the same with regards to "accurate" colour, so which one should I buy? Regarding the poster's question then if he still wants to pursue it then importing the image into PS and searching for a mid grey with the colour picker and averaging the result by moving the colour channels in curves will neutralise any cast, but is it worth it?
-
Help me out with some color management basics;
As far as I understand most systems are calibrated for some standard light, say daylight or flash, 5500K or so. With that kind of illuminant I guess it's possible to get accurate colors out of most cameras with appropriate profiles
Hi,
Yes, within the limitations of a tri-chromatic color sampling model. There will always be trade-offs in accuracy, and there will be issues with transitions between the three bands. It's also important to remember that a color temperature represents a continuous spectrum of light, one that resembles the emission of a so-called 'black-body' emitter of said temperature in Kelvin.
The other complicating factor is with the subject colors, some of which may be metameres which may look different depending on the spectrum of the illuminant.
However, what happens if you shoot in the dusk or dawn, or like me in the winter above the arctic circle, the color temperature of the light is then far from the standard 5500 or 6500K. I have a hunch that this affects how the eye/brain experience colors so you would need some other color model for these temperatures in order to make a calibration to accurately support that. Is there such a thing, or is accurate color management locked to illuminants close to 5500K?
As long as we consider pure color (continuous/incandescent light, pigments, dyes), changing the whitepoint will be something that can be handled mathematically quite well, with some chromatic adaptations. But as soon as we introduce metameric color or filtered emission spectra, things get a bit less accurate. When high accuracy is needed, it becomes inevitable to calibrate for those lighting circumstances, but metamerism and a tri-chromatic color sampling will continue to be an issue.
Why would I want to have accurate color in extreme color temperatures you ask? If I could I would, I love to have "neutral" as a starting point when I process my pictures, but I have noted that for these light conditions it's not easy. I'd just like to know if it's possible with current technology and color models.
I also like to get a neutral starting point. Not to eliminate e.g. the colors of a sunset, but to eliminate e.g. strange magenta aberrations which are caused by our imperfect color models, before I dial some color temperature shift back in.
There are also vision related complications, e.g. due to absolute illumination levels, and the way we see simultaneous contrast and color. That's something that a more perceptual CIECAM model can address better than a simplified color temperature and tint model can. Cliff Rames published (https://sites.google.com/site/clifframes/ciecam02plugin) a Photoshop plug-in which allows to experiment with that. RawTherapee (http://rawtherapee.com/downloads) supports some CIECAM functionality as well.
So, it won't be possible to get it accurate due to the limitations of our tri-chromatic sampling and spectral discontinuities and imperfections, and how human vision functions, but we should be able and get within the ball-park. If necessary one should profile for the actual situation, but creating accurate profiles is not that simple either, especially outside of a controlled studio environment.
Cheers,
Bart
-
In extreme color temperatures I think that the regular "daylight" white balance often produces a quite realistic result, as it will simulate the effect when the brain does not fully compensate for the extreme color temperature (ie something that appears white under daylight does not look white any longer).
Setting the white balance from a grey card won't work, because in these extreme temperatures the eye/brain does not fully adapt and I can see that the grey card looks blueish or reddish or whatever.
I've attached an image from a challenging situation, it's shot during the winter some time after sunset with a 5D mark 2. The light was extremely soft with very low contrast, and the eye did not experience the snow as "white", but as colored by the colored sky. The images from this shooting session was very hard to get right so it looked like my eye's impression on the scene. I tried various profiles from various converters, but I ended up using standard color matrixing because it more closely resembled my memory of the colors. I shot over a few hours and all the time the light was changing as the color of the sky was changing, so a shot made just 40 minutes earlier had quite different color.
If I had not processed the images so closely after I shot them, it would have been hard to make the proper color decisions. I'm quite sure that I got this image closely resembling the "reality", but some other images were much harder.
In these situations I'd prefer to have some better color management method than "process as soon as you can after the shot, and try to recreate from memory by trying various profiles and settings".
-
by the way: I'm totally aware of the extreme challenges of mixed artificial light, narrow band lights being the worst. But I thought that maybe for the broad band before/after sunset sky light there could maybe exist color models that could work.
-
Setting the white balance from a grey card won't work, because in these extreme temperatures the eye/brain does not fully adapt and I can see that the grey card looks blueish or reddish or whatever.
...
In these situations I'd prefer to have some better color management method than "process as soon as you can after the shot, and try to recreate from memory by trying various profiles and settings".
It is my understanding that "color profiles" effectively attempts to relate the non-standard spectral response of a given camera to something more like a (standardized model) of our eyes - valid ideally for one set of illuminants (realistically more if both illumination and camera spectral sensitivity is smooth). Assuming that the spectral response of our photoreceptors is constant, one might hope that this is a good thing for any scene where you wish to recreate human perception?
While White-balance conceptually is a subsequent process that attempts to "replace" the illumination of the source scene with that of the destination scene. For scenes where the illumination is part of the joy, and not something that you want to get rid of, the WB concept seems kind of redundant.
To really recreate such a look, you would need a self-radiating image (such as an LCD monitor), not a reflecting print, would you not (furthering the though of recording and recreating the spectral distribution with little perceptual adjustment)?
-h
-
Sorry,
I was really responding to the subject line and not the entire posting. I'll try to read the entire posting next time.
Best regards
Erik
by the way: I'm totally aware of the extreme challenges of mixed artificial light, narrow band lights being the worst. But I thought that maybe for the broad band before/after sunset sky light there could maybe exist color models that could work.
-
Setting the white balance from a grey card won't work, because in these extreme temperatures the eye/brain does not fully adapt and I can see that the grey card looks blueish or reddish or whatever.
The problem is making a choice of which "accurate" is the one you want. What your eye can see is one definition of accurate, but that's different than what can be measured, and both are different from what may look best on a photograph. To record what you saw does require processing while the memory is fresh.
At lower latitudes the significance is not as obvious. "High noon" differs from mid-afternoon, and the Golden Hour lasts about an hour. It's not so simple at high latitudes. In June my Golden Hour takes up maybe as much as 8 to 10 hours. It's the mid-day period that is short! In December none of that happens, and the bluish light of shade in failing civil twilight is all there ever is. We get approximately the same light as everyone else about twice a year, at the equninox.
I want the "best" color for a dramatic photograph. I might process half a dozen consequtive shots to have a different color of sky behind the same bird, as an example. I want the bird to stand out, and if different contrasts or color balance help with that, so be it. That is very different than what you are targeting.
By the same token, your attempt at what you saw and my attempt at what I think is pretty are no better than the next guy wanting whites that are white, reds that are the same red everywhere, and so on.
All three types of "accurate" become more technically complex under more variations in light. Starting at sub-Arctic latitudes that begins to become very significant and is increasingly so above 64 degrees. There are no simple solutions.
-
Sorry
I was not reacting to that :), I said that to try to narrow down the discussion to a case that might(?) be solvable.
-
There are no simple solutions.
That was what I was guessing :). In technical terms that the experiments that have been made and standardized to create the color models we have do not include how the brain/eye experiences "extreme" color temperatures.
It probably would be possible to make such experiments and derive a color model for various types of lights, so you could when out in the field have some sort of spectrophotometer to measure the ambient light more precisely and apply the appropriate color model. Would still not work in mixed light, but could work in the example picture I posted earlier in this thread.
From an artistic standpoint I'm kind of getting bored of the typical saying "I have printed this landscape photograph not to represent what I saw, but what I felt", to me that's just another way of saying "I have manipulated this picture to make it look like I would have liked it to look". I have nothing against those that have this style, landscape photography has no documentary rules. But for myself I'd like to try a new concept, having colors that actually just as "dull" and toned down as in real life, I've figured out that since noone is doing it, it would be a pretty unique style ;-)
The ability to modify color today is so high that I feel that when I see fantastic light and color in real life it's not as exciting to capture it as I know a similar popping color could be post-processed from a duller base. However if I really had the ability to capture true-to-the-eye colors under most conditions I could have my own set of "dogma" rules that limits me how much I'm allowed to push processing.
-
I have to agree with the general gist of Floyd's arguement.
For colour to be accurate one has to have a reproducible standard to compare it to.
Then there has to be a compelling reason to make the colour reproduction accurate (to the reproducible standard).
In the case of the posted jpeg "correcting" the colour to some supposed standard is nonsensical because the entire aesthetic of the image is based on the fact that parts of the scene that colour correction might render white or a shade of gray are in fact beautiful delicate pinks, blues, and purples.
Perhaps there really is some standard that will allow one to colour correct an image like that but it is likely that the result would look truly horrid and artificial.
If one is doing product or food photography then shooting in a controlled environment with light of predictable spectral characteristics and doing colour corrections in post is a no brainer.
Shooting outdoors one is looking for unusual light that, either for practical or aesthetic reasons, should not or cannot be corrected because the result will look either boring or horrid.
Best to rely on your eye and your memory in these cases.
My $0.02 worth
Tony Jay
-
I think there may be a confusion of what I mean by "correct" or "accurate" color. Forgive me I'm new to color management, but to me I see the eye/brain as the reference, ie correct color for extreme color temperatures should be how the eye/brain responds in that light, not how the objects would have appeared in noon daylight.
In other words, if I shoot a color checker in that dusk arctic light and make a profile that renders the white as white etc I don't consider that as "accurate" color as you're not representing how the eye/brain saw the original scene. The eye/brain only "white balances" within a reasonable range of color temperatures, at extreme temperatures it can even out some but not totally white balance, eg the snow is no longer white as it was at midday.
If I were a color standardization organization I suppose I could measure some different lights with a photospectrometer, recreate that under lab conditions, gather some test persons and have them do color matching tests so one could standardize how the eye/brain experiences color under certain lighting conditions. Maybe these tests would be plagued by all sorts of metamerism issues, I don't know.
My original question was if there already is such a thing, ie if there are tools today that allow you in a reasonably accurate way render colors how the eye/brain experienced it at the scene for "extreme" color temperatures, or if the way you have to do it is to try remember how it looked like and then recreate as good as you can in your raw converter.
Maybe if I just could set an "off-white" balance it could bring me far, that's generally how I do it now manually. I white balance the picture but let the white have the color cast I remember it had, usually by trial and error of color temperatures and tints until it looks sort of right. But sometimes I just feel lost. "Is this too blue? Or too little blue?" etc..., especially when there's been some time from the shoot until I process.
-
Given that you experienced a certain visual impression that day.
If somehow, a wide-angle image could be recorded where at each pixel, a highly accurate estimate of the light intensity for each of a set of narrow passbands (e.g. 10nm from 400nm to 800nm) and recreated. Given that the recreated image was rendered filling your entire field of view (using a large curved projector screen or some goggles). I believe that this might be possible, if not practical, by using a colour-wheel and multispectral imaging/DLP projection along with some tinkering for stationary objects.
Would we not expect to be able to recreate that same visual impression? Excluding the inevitable "subjective" factors of freezing your behind off, being on a skiing trip etc.
If the scene can be recreated by the (impractical) means above, what does it say about the available (practical) solutions that fails to do the same?
-h
-
The fact that you are asking so many questions about this issue should mean that in reality it isn't possible?. Concentrate on what is "pleasing". At the end of the day if you colour correct an image and it pleases you you should then be happy. If you show it to someone then they won't have a clue about how the original scene looked like and they won't care?
-
If there is a tool I am not aware of it.
As I mentioned before your eye and your memory are actually the best guide in this situation.
In the terminology that you are using the "correct" colour is the colour that is most aesthetically pleasing to you in the context of the overall image.
The fact that another photographer who was present might develop your image differently in post (with regards to colour) is neither here nor there. Aesthetics will be the goal.
You may favour somewhat more subdued colour, perhaps warmer hues or colder. It is your call.
There will be many interpretations that will look natural and believable to a third party.
The one that looks best to you will likely be the one that eye and memory informs is closest to what you experienced.
Tony Jay
-
Thanks for all your input.
There sure is a psychological component to color, so it's easy to get to the conclusion "why care?". People that work with artwork reproduction photography don't have this relaxed relation to color. I'm asking around if there are methods to be as anal about color as they are when you're a landscape photographer. My initial guess was that the answer is "no", and that seems to be correct. Tough luck, I'll just continue guess the best I can as I do today.
If there were methods, I'd probably use them. Why? There are two reasons. I think photography becomes more rewarding to me if I can capture and render color close to how the eye experienced it at the scene, as it makes special light more special rather than just a post-processing product. The other reason is that I often enough shoot in extreme light (arctic winter really is special) and have time pass between shooting and processing so I just get lost when post-processing and would like to have a sane starting-point.
Attached another photo from the same occasion as the first posted earlier. This was shot a little bit earlier, the mountain to the left is the same as in the first picture. The light is similar but quite different, the sun is still just above the horizon. I remember that for this photo I had quite a lot of difficulties to render color that felt true to what I had seen, it was easier with the photos shot just 30-40 minutes later when the sun was down.
-
The fact that you are asking so many questions about this issue should mean that in reality it isn't possible?. Concentrate on what is "pleasing".
Frankly, that is IMHO a bit evasive. It may not be possible to do it exactly accurate, but we may still be able to do it somewhat accurate. We know that our camera, even under optimal conditions, doesn't create an exactly accurate color image. If it did, then color temperature would be exactly the same as the (incandescent) illuminant emission, and tint would be zero. And even then, metameric color may look different than it does under a different color temperature, and ambient reflections may create a local color cast.
So, what would be useful is to be able and eliminate the obvious tint errors, as much as possible, if they can be attributed to the illuminant. Then, and then only, can we adjust towards 'pleasing' rendering (whatever that may be for an individual observer), also taking local ambient color reflections into account.
The real challenge is in determining the correct White balance for the main illuminant, usually with only our inaccurate tri-chromatic sampling device/camera to measure it with. We either have to rely on reflection from objects in the scene, which may fail due to non-uniform spectral reflection, or on the main light source which may be filtered or has a deviating emission spectrum.
At the end of the day if you colour correct an image and it pleases you you should then be happy.
That seems a bit of a circular reasoning to me, the colors are pleasing when they are made pleasing. The OP was wondering if there is a more objective way to arrive at that goal. Since there may be many pleasing renderings possible (also depends on (lack of) taste), why not pick one that is at least somewhat accurate?
I think that (if possible) a WhiBal or ColorChecker or other spectrally neutral reference object that is directly illuminated by the main illuminant (or its reflecton, e.g. blue sky in shadows) can help to eliminate most of the (tint) inaccuracies thus leaving our sense of taste or creativity to give it a spin which is at least founded on a stable/believable starting position.
When I want to create e.g. a pleasing sunset image, I know that the balance between shorter and longer wavelengths is distorted in a somewhat predictable manner. I do not want other color casts to dominate that image if my goal is to approach a somewhat accurate representation of that scene. From there it is still possible to adjust saturation or color temperature, but then that won't boost inaccurate tint influences. It basically takes part of the guesswork out of the equation.
Cheers,
Bart
-
The problem is making a choice of which "accurate" is the one you want. What your eye can see is one definition of accurate, but that's different than what can be measured, and both are different from what may look best on a photograph.
Exactly. We can use the term accurate and colorimetry (the measurement of color) together and a room filled with people viewing the results can say it doesn't look 'accurate'. The term accurate has to be defined. It has to be ignored as a color management sales and marketing buzz word.
-
Exactly. We can use the term accurate and colorimetry (the measurement of color) together and a room filled with people viewing the results can say it doesn't look 'accurate'. The term accurate has to be defined. It has to be ignored as a color management sales and marketing buzz word.
It seems like you are avoiding or missing the central question. I guess it was a mistake by me to use the word accuracy, as it draws attention from my question.
Let me rephrase. There are color models and methods that try to achieve some sort of well-defined color reproduction in controlled situations, used in for example art reproduction. When you photograph a painting for posterity you don't want it to be some personal taste of the photographer that decides how the colors are rendered. Instead you want to have some standard based on experiments on the color-seeing population so you will render something that is close to the colors most people see when they look at the original painting. You're an expert at this so you know all about it.
Now, what I'm asking is if there are similar or alternate methods to use when working in extreme color temperatures such as when shooting an outdoor landscape in arctic winter in dusk or dawn. A challenge here is that the color temperature is so extreme that the eye/brain does not compensate fully, the once white snow is no longer white, thus if I want to reproduce a photograph as realistically as possible I should not fully whitebalance the snow, but let it be off-white in the same way as people experienced it at the scene. As far as I know the only way to achieve this with today's tools is to use your memory and guess, no color management tools exist to help you choose a reasonable cast for your white to make your photograph be a realistic representation of the eye's experience at the scene.
From this discussion it seems that my assumption was right -- that there are no such methods, and all color models which strive for some sort of accuracy (or realism perhaps is a better word) work in a fairly narrow color temperature range. Ie, landscape photographers working in dusk or dawn and wanting realism must trust their good color memory and post-processing skills. So I'm not missing out on some color management method I as a landscape photographer should know about, which was my main concern and the reason I opened this thread.
-
Let me rephrase. There are color models and methods that try to achieve some sort of well-defined color reproduction in controlled situations, used in for example art reproduction.
Key word in the above sentence is try. Trying and achieving (to a hopefully well defined metric anyone can agree upon) are two different things! And what color models do you refer to and do they succeed in providing the match without massive efforts by some human to produce what someone says is a match?
When you photograph a painting for posterity you don't want it to be some personal taste of the photographer that decides how the colors are rendered. Instead you want to have some standard based on experiments on the color-seeing population so you will render something that is close to the colors most people see when they look at the original painting.
What standard are you referring to?
From this discussion it seems that my assumption was right -- that there are no such methods, and all color models which strive for some sort of accuracy (or realism perhaps is a better word) work in a fairly narrow color temperature range.
Yes and no (more yes than no). Today's color management makes a lot of assumptions that may not pan out and produce what you would like to consider as accurate color. A few examples: Nearly all ICC profiles assume a D50 viewing condition when that may not be the case. Or that only one object in our solar system is capable of producing D50 and further, D50 is a sampling of colors taken all over the planet. Or that the current models used for color management have plenty of issues that produce color matching problems (example, Lab's attempt to create a a perceptually uniform color space which isn't perceptually uniform, or how it exaggerates the distance in yellows and consequently underestimate the distances in blues) again producing color matching issues. Or that we have conditions where metameric failure occurs, even with the observer. That there are color appearance models still in the works and not implemented in current color management solutions. That our capture systems can capture 'color' we can't see or we can see colors it can't capture. Etc, etc, etc.
Ie, landscape photographers working in dusk or dawn and wanting realism must trust their good color memory and post-processing skills. So I'm not missing out on some color management method I as a landscape photographer should know about, which was my main concern and the reason I opened this thread.
My advise is to continue making lovely images and don't get caught up in an accuracy, color matching rabbit hole. You shot film once? Was that an issue? Because none of the films reproduced the scene as it appeared to you and they all had renderings that were built to convince you to buy that film product. Digital isn't much different. We have large numbers of users who can't get a decent print to screen match let alone reproduce a single print that is accurate to the scene. An emissive display and reflective print will never match perfectly. To extrapolate past that 'issue' and ask, how can I produce a capture, then a print that is accurate to the scene seems awfully difficult and further, is it even necessary?
-
I'm a little bit familiar with color management but no expert so it's risky for me to point out any standard in particular. Let's say there are various of organizations and companies that work with color and try to solve various challenges related to accuracy, and different standards and methods has arised, to calibrate cameras, screens, printers, link them together in a color managed workflow, with ICC DCPs color checkers, spectrophotometers and the like. I have myself a color managed workflow from screen to print. I know color management is about reducing the errors not eliminating them, because you can't.
I'm not planning to get caught up on anything, just checking what the possibilities are. If it was possible to make something more repeatable than me guesstimating white balance and individual color adjustments from shot to shot, I'd love to use it. It would speed up my workflow instead of having white balance picking angst. It would also contribute to make my images have a more similar and stable look from year to year instead of guesstimating freshly each time.
There sure are impossible special cases, the northen lights for example, are recorded very different on camera than with the eyes. And narrow band mixed artificial lights is also an impossible case. I just thought that there might be a chance for better methods for dusk and dawn light.
The reason it does not exist I don't think is because there's no need or interest, but because it's hard, impractical or perhaps even impossible to make anything better than the photographer randomly pulling the temp and tint slider, which indeed seems to be the established method :).
-
Let's say there are various of organizations and companies that work with color and try to solve various challenges related to accuracy, and different standards and methods has arised, to calibrate cameras, screens, printers, link them together in a color managed workflow, with ICC DCPs color checkers, spectrophotometers and the like. I have myself a color managed workflow from screen to print. I know color management is about reducing the errors not eliminating them, because you can't.
The bottom line, and the reason I posted was to let you know that the term accuracy, in relationship with color management is mostly a marketing term and without some metric to what accuracy goal you desire, you going to get caught up with semantics or far worse, marketing semantics.
If you tell me you want to make a print today and one in a year and you desire that the printer exhibit no more than a Avg dE2000 value of 1, AND you define how many patches you wish to use such we can have a firm idea of what the average represents, that's very doable with todays tools and software. If you tell me you want your capture and a print to be accurate, I can try to sell you all kinds of products and blow smoke up your behind which doesn't serve you well.
If it was possible to make something more repeatable than me guesstimating white balance and individual color adjustments from shot to shot, I'd love to use it.
Yes, you could setup a test lab, capture known color values today, analyze the numbers and test that in a week to see the dE differences. How that will aid you in the field is highly questionable. And even if we learned that your camera has a dE2000 difference of X number of colors shot today and in a week, due to say temperature or something else, now what? It is one thing to be doing astrophotography or some laboratory capture process where the dE values have to be lower, but since that isn't what you're going to do with the device, but instead make pleasing images, is this useful, can you even do anything about this behavior?
It would speed up my workflow instead of having white balance picking angst.
Get something like a Passport or similar target and photograph it before each capture. Now you come back from shooting a sunset and guess what result you'll get if you white balance? Probably not what you desire and certainly not accurate color to the scene (scene colorimetry). In the studio under the same strobes: that could be a very useful tool and workflow.
It would also contribute to make my images have a more similar and stable look from year to year instead of guesstimating freshly each time.
Yes, if you always captured in a controlled studio condition with the same strobes. If you are doing product photography, that be quite useful!
There sure are impossible special cases, the northen lights for example, are recorded very different on camera than with the eyes.
Begging the question, do you want the image to appear as you thought you saw it and thus probably inaccuracy or as the camera saw it which could also be inaccurate due to the limitations or idiosyncrasies of the device and not what you wish to represent in your image?
The reason it does not exist I don't think is because there's no need or interest, but because it's hard, impractical or perhaps even impossible to make anything better than the photographer randomly pulling the temp and tint slider, which indeed seems to be the established method :).
I'd say both and I'd say it highly depends on what you're doing with a camera. If you're photographing cytology under a microscope and hoping to show others what you saw by sending them an image, that's one thing. If you're shooting then northern lights and your goal is to impress your photographic vision or what you remembered you saw, that's a different story.
-
I've mentioned several times in this thread that using a color checker as reference won't work, as in extreme color temperatures, well even moderately off 5500K the eye/brain does not fully compensate and starts to see a cast on objects that look white under 5500K.
I've attached a very clear example of this. Snow is experienced as (quite close to) pure white under midday. The attached picture shows the same snow and ice shot at the same occassion, with minutes apart, but one has a "creative" white balance, ie the snow is used as white reference, while the other I have manually tuned from my best ability to make it a realistic representation of what the eye saw at the scene. The result would be very similar if I white balanced with a color checker or grey card, ie very very far from realistic. Sure you can make a pleasing image, but I also like to be able to make something realistic, preferably in a structured repeatable way.
Today I don't work so much with pulling individual colors. I try to remember what the cast of white was, or how pink the sky was etc, and then I pull the white balance to try to get a reasonable match with memory. Then I have a couple of DCP/ICC profiles (made for 5500K I guess) from different software makers and the plain 5500/6500 color matrix (which I think adobe has come up with) and I do trial and error with those to see if it becomes better or worse. Sometimes the color matrices gives me better result than the profiles. When I shoot with the DSLR I sometimes tune the white balance on the camera so the LCD image looks similar to what I see, so I have some sort of reference embedded in the RAW to support my memory.
Relying on my memory and pull sliders to match is the least bad method so far, and now I know that there probably are no better methods that could help me to get a realistic color rendering baseline. Tough luck. I want to capture and show the unique beautiful light as it is up here in northern Sweden (you don't see this kind of dusk light even in the southern parts of Sweden, the light becomes extremely pink/blue), but that only makes sense if I'm able to render it realistically. If I just do it on chance or as I please I could have shot it in much different light and just tune it. So I don't think it's strange to desire realism, but I understand that with current technology and understanding brain/eye function in these type of light conditions there are not many tools for the photographer to use.
-
I've mentioned several times in this thread that using a color checker as reference won't work, as in extreme color temperatures, well even moderately off 5500K the eye/brain does not fully compensate and starts to see a cast on objects that look white under 5500K.
The Passport has differing colored whites (cool to warm) so you can season to taste which isn't accurate color (it's pleasing color). You're shooting raw right? The WB settings have zero affect on that data, you can render it anyway you desire.
I've attached a very clear example of this. Snow is experienced as (quite close to) pure white under midday.
Visually or numerically or both? What numbers, the actual scene referred values or what you end up with after altering the settings?
The attached picture shows the same snow and ice shot at the same occassion, with minutes apart, but one has a "creative" white balance, ie the snow is used as white reference, while the other I have manually tuned from my best ability to make it a realistic representation of what the eye saw at the scene.
I would submit that unless you measured the illuminant at the scene, both are creative, you apparently prefer one over the other. Again, unless you can measure the colors at the scene and of the final, there's no actual colorimetry happening and everything else is just a set of numbers you either prefer or you don't. That's subjective.
Sure you can make a pleasing image, but I also like to be able to make something realistic, preferably in a structured repeatable way.
Tell us what the differences are? What makes one that you say is realistic and accurate instead pleasing?
I try to remember what the cast of white was, or how pink the sky was etc and then I pull the white balance to try to get a reasonable match with memory.
How is that accurate? What metric can you use to suggest it's accurate since there's nothing measured?
Sometimes the color matrices gives me better result than the profiles.
Better as preferred. That's subjective.
What I'm suggesting you to do is consider what you're asking for (accurate color) and how to get it. What you're describing is a desire for a pleasing rendering. You can't record what you remember of the scene inside your brain. What you ate for breakfast could affect your perception that morning, before we even get into the impossibility of defining memory color. Calibrate and profile your display, adjust to taste to produce a rendering you are pleased with that you and only you feel represents the color you recall. I understand you want this to happen more automatically but short of capturing a reference target as a start, or having a few decent camera profiles as a starting point, there isn't much more you can do. A large part of Photography is rendering a scene you wish to represent it to your viewer: http://wwwimages.adobe.com/www.adobe.com/products/photoshop/family/prophotographer/pdfs/pscs3_renderprint.pdf
Read this piece (http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf) and examine the dark and unattractive Figure 1. That's vastly more accurate than the other two figures. It's scene referred. But does anyone prefer that rendering over the other two images where are output referred and numerically far less accurate? Doubt it.
-
Yes I shoot raw :-). The camera rendering still get embedded as a preview image for most raw formats.
As you can see from the images I attached the color differences are not subtle, they are huge. So yes there is no doubt that my manual rendering is more realistic than the snowy white balance, regardless what I had for breakfast. I had a fellow photographer with me and her renderings look similar to mine - although there is a lot of psychology involved trained healthy eyes get to the same ballpark. I know memory certainly is unreliable, but what to do? As you say there's not much there is. It does not seem attractive to me to replace an inconsistent 'okay' (my memory) with a consistent bad (color checker). I'll think I'll give the checker another chance though, the reference could be used for at least something.
When it comes to dullness of "accurate" renderings I know about that, but it's still a good starting point for processing, and often a contrast curve and slight saturation increase is all that is needed. Being able to get the hues reasonably realistic is my most desired property.
-
Yes I shoot raw :-). The camera rendering still get embedded as a preview image for most raw formats.
You mean the proprietary JPEG rendering that the camera provides that has no direct bearing on the raw data itself?
As you can see from the images I attached the color differences are not subtle, they are huge.
The are but they don't have to be. That's why you capture raw data.
So yes there is no doubt that my manual rendering is more realistic than the snowy white balance, regardless what I had for breakfast.
No argument other than your manual rendering is more to your preference and accuracy can't be attached to this without first having some measurements of the scene. Measurements that will not match what you produced because you produced an output referred image.
I had a fellow photographer with me and her renderings look similar to mine - although there is a lot of psychology involved trained healthy eyes get to the same ballpark.
And that's not a surprise but it also doesn't have any direct correlation to accuracy.
I know memory certainly is unreliable, but what to do? As you say there's not much there is. It does not seem attractive to me to replace an inconsistent 'okay' (my memory) with a consistent bad (color checker). I'll think I'll give the checker another chance though, the reference could be used for at least something.
IF the addition of a color checker doesn't help, don't use it. You could just use that in-camera JPEG which may have no relationship to reality in terms of scene accuracy as a starting point. That is again just another rendering interpretation that isn't any more accurate. Just as if you captured the scene with Velvia, Ektachrome and Agfachrome, all would be a different rendering. You'll prefer one. None are accurate. You pick a film stock based on the rendering you prefer.
-
Torger,
I am not a color scientist, but I share your view of the problem (as well as the love for Lappland :-) ). May I humbly point you to an attempt I have described here:
http://www.luminous-landscape.com/forum/index.php?topic=73620.0
I have not yet had the time to experiment with it in more detail. It might be interesting to hear if/how it works for you, who has daily access to "extreme" white balance situations.
Good light - and "true" color!
-
Help me out with some color management basics;
As far as I understand most systems are calibrated for some standard light
consider that bayer CFA is on top of a particular sensor surface is set once and for all when camera is designed and manufactured and CFA is designed w/ specific properties of "R"-"G1-"B"-"G2" filters... it is reasonable to assume that you can design a "better" set of filters for a specific spectrum of incoming light, but you can't alter CFA properties on the fly to account for different types of illumination... so yes, it is designed for something like "daylight" and it is naive to think that post processing can fix that 100%... "damage" (metameric failures, etc) is already done before any software or profiles come into picture.
-
Torger,
I am not a color scientist, but I share your view of the problem (as well as the love for Lappland :-) ). May I humbly point you to an attempt I have described here:
http://www.luminous-landscape.com/forum/index.php?topic=73620.0
I have not yet had the time to experiment with it in more detail. It might be interesting to hear if/how it works for you, who has daily access to "extreme" white balance situations.
Good light - and "true" color!
I see that you use the same support technique as I do, estimating the white balancing by matching the LCD/live view, just a little bit more elaborate than I've done so far. I also see that you get the expected critique; that the screen is poorly tuned sRGB etc, but we all know that has worked with this that the casts of screens (with a few exceptions, use a good camera model!) and matching errors is much less than the range of white balances you can be confused about in post when you work with these type of scenes and have no recorded reference.
One has to differ between small errors and large errors. Some seem to think as there will always be small errors (and they are good at pointing out those) you could just as well skip it all-together and have very large errors, and then go on questioning why one would want realistic color at all. I don't think that's a very helpful approach. Some of us maybe want to have realism as a style, just as others have surrealism. And even if I want surrealism, at least I prefer to have a realistic starting point so I know what I'm doing. So I'm glad to see that there's someone else that have this problem and have found a technique that makes the post-processing challenge a bit easier.
I like Schewe's anecdote in your thread, about the lecture of white balance in antarctic light. Polar light is special and you get an unusual white balance challenge. The lecturer's solution was to do the "from memory" thing until it looks right, but I think your method by manually tuning and screen matching on site is providing good support to find a fairly realistic starting point in post. The possibility to do so is fairly new, as one needs high quality live view renderings so it's certainly a technique worth evaluating further.
-
The physical equivalent of a "color picker" sounds like a neat thing. Use a modified colorimeter or some other gadget, for either very small or very large AOV. Point it towards the sky, the land or the sun. Get a reading. Use it afterwards as a reference point.
These things are commonly >3 bands, are they not?
-h
-
The physical equivalent of a "color picker" sounds like a neat thing. Use a modified colorimeter or some other gadget, for either very small or very large AOV. Point it towards the sky, the land or the sun. Get a reading. Use it afterwards as a reference point.
These things are commonly >3 bands, are they not?
Pro spectrophotometers are very expensive of course ($10K++), but a consumer model like my Colormunki is like $500 and measures in 10nm bands. With optics it could possibly be a spot meter style color picker. Trouble is what we should do with those values, how we should get them into the workflow so it translates to more realistic and repeatable color reproduction. Maybe a full spectrum reading of a gray card would be easier to use and provide enough data to make something better than current guesstimating techniques, but we would need to develop color models that translates "extreme" spectrums into some sort of white balance. The current models that exist don't deal with extreme spectrums as far as I know.
I'm suspecting that a live view camera with good display and manual matching at site would be a simpler technique which would give about the same size of the error. This method is exactly the same as the established "pull sliders randomly in post until it looks like you remember it", but you do it on site with the real scene as guide.
I know the method is open for lots of criticism concerning calibration of the camera display and the very strange viewing conditions, but fact is that the range of white balance confusion ("should I use this or that?") that can arise for raw files shot under say arctic dusk light is very large, and as far as I can see much larger than errors introduced with the matching method. When I work with post-processing I use a calibrated screen, but I have noted that even when looking at images on an uncalibrated screen you see the same types of color errors (tints) you do with the calibrated screen, ie calibration is only necessary for really fine-tuned work and print matching, the errors we talk about here when trying to represent extreme landscape light in a realistic manner is on a much grander scale. Therefore I think it could be a quite good method actually, but I need to evaluate it more, and winter is coming :-).
-
torger you seem very knowledgeable about the subject which makes me wonder why you posed the question in the first place?
-
Hi,
My guess is that it is best to find a color temperature and tint that works.
Part of the problem is that you have a mix of blue skylight and yellow-red sunlight, both add to perception.
Best regards
Erik
Pro spectrophotometers are very expensive of course ($10K++), but a consumer model like my Colormunki is like $500 and measures in 10nm bands. With optics it could possibly be a spot meter style color picker. Trouble is what we should do with those values, how we should get them into the workflow so it translates to more realistic and repeatable color reproduction. Maybe a full spectrum reading of a gray card would be easier to use and provide enough data to make something better than current guesstimating techniques, but we would need to develop color models that translates "extreme" spectrums into some sort of white balance. The current models that exist don't deal with extreme spectrums as far as I know.
I'm suspecting that a live view camera with good display and manual matching at site would be a simpler technique which would give about the same size of the error. This method is exactly the same as the established "pull sliders randomly in post until it looks like you remember it", but you do it on site with the real scene as guide.
I know the method is open for lots of criticism concerning calibration of the camera display and the very strange viewing conditions, but fact is that the range of white balance confusion ("should I use this or that?") that can arise for raw files shot under say arctic dusk light is very large, and as far as I can see much larger than errors introduced with the matching method. When I work with post-processing I use a calibrated screen, but I have noted that even when looking at images on an uncalibrated screen you see the same types of color errors (tints) you do with the calibrated screen, ie calibration is only necessary for really fine-tuned work and print matching, the errors we talk about here when trying to represent extreme landscape light in a realistic manner is on a much grander scale. Therefore I think it could be a quite good method actually, but I need to evaluate it more, and winter is coming :-).
-
Pro spectrophotometers are very expensive of course ($10K++), but a consumer model like my Colormunki is like $500 and measures in 10nm bands. With optics it could possibly be a spot meter style color picker. Trouble is what we should do with those values, how we should get them into the workflow so it translates to more realistic and repeatable color reproduction.
If you point a colorimeter (or spectrophotometer) converted to spot-use towards some point in the sky, you would get a precise measurement of the spectral characteristics of that point, yes? (If you had a lot of spare-time, you could raster-scan your scene like an old video camera, and get a spatially crude, but spectrally accurate representation)
Print the image, hang it up on the wall under some lighting, and do a second reading of that same spot (if the sky is smooth, spatial accuracy may not be that important). Or trust the profiles of this printer to be "accurate".
You wish for those two readings to be similar, right? So what operation is needed on the image data to make the print have the closest possible spectral characteristics at that point to those of the original scene (within the degrees of freedom in the ink, paper, lighting etc)? Many operations would likely get you there, but you'd want one that does not mess with tonality, perhaps a simple 3x3 color channel matrix.
What I am suggesting (without really knowing what I am talking about) is a poor-mans substitute for a multispectral camera, rendered with "absolute" spectral response in mind. If your sky had a spectral peak at 620nm and a valley at 640nm etc and you want to recreate just that, then why mess around with concepts such as white balance, illuminant, human perception etc any more than necessary? If for nothing else, if would be really interesting to see if a "camera color system designed by a physics academic with little interest in human perception" could capture your scene more to your liking.
My working hypothesis is all along that if you can recreate all of the (perceptually relevant) physics stimuli that caused such an impression in the real scene, then your vision should respond the same way if presented to the same stimuli at a later time. Generally, this assumes that you cover your entire FOV and that there is no coupling between eyes, ears, taste, knowledge etc. In the abscence of anything better, I think that is fair.
-h
-
Generally, this assumes that you cover your entire FOV and that there is no coupling between eyes, ears, taste, knowledge etc. In the abscence of anything better, I think that is fair.
And you forgot about having a good memory....especially if you took about 100 shots in that particular day? ;)
-
And you forgot about having a good memory....especially if you took about 100 shots in that particular day? ;)
I am curious as to why these problems appear, in such cases I have been known to offer suggestions and thought-experiments that can be impractical, investigate asymptotic behaviour etc.
Would a multi-spectral camera (if one suitable were available) coupled to a multi-spectral workflow and printer solve the problem? Or is it more philosophical about the idea that one can do a 2-d "image" of a scene, move it to another time another place and expect to recreate the same perceptual response?
-h
-
I am curious as to why these problems appear, in such cases I have been known to offer suggestions and thought-experiments that can be impractical, investigate asymptotic behaviour etc.
Would a multi-spectral camera (if one suitable were available) coupled to a multi-spectral workflow and printer solve the problem? Or is it more philosophical about the idea that one can do a 2-d "image" of a scene, move it to another time another place and expect to recreate the same perceptual response?
One problem is the brain's evening-out-effect, most clearly demonstrated with mixed artificial light. If you photograph that and want to have a realistic representation you will feel that you need multiple white balances, to bring the lights closer together as it was experienced on scene.
The simplest way to see on this is that the brain fully compensates, ie "auto white balance" so the brightest spot is seen as pure white, and to model this behavior one can shoot a gray card and the white balance with a color picker. However it's more correct to say that the brain brings the "neutral" color closer to pure white but not all the way. In arctic dusk light the "white" is experienced as very far from white so this model will not yield a realistic result.
As far as I know there are no models made of the brain's "white balance" function, so even if you have a full spectral recording you don't have a model that can translate that into how the eye experienced it. I do think it would be possible to make such a model based on experiments. Another aspect is that possibly the color response changes more drastically (like it does in low light) than just a white balance difference so you would need a totally separate DCP/ICC for more extreme light. The standard DCPs are typically made for 5500 and 6500K. I don't really know how well those translate into extreme light. Probably not perfectly, but my guess is that it's not too bad either, if you can secure the white balance then you have come a far way.
The eye/brain has other "evening out" effects to, for example that you much easier can see past reflections on a window (it's about depth seeing) than in a photograph, or that you don't see haze as disturbing in real life as in a photograph, which means that you may use a polarizer to reduce reflections and slightly reduce blue channel content on distant object to get more realism.
And then we have all other sorts of lighting conditions, low light, narrow band atmospheric phenomena like northern lights etc.
(And we have print paper and viewing conditions which affect the experience, but I think people have a tendency to emphasize these issues way too much compared to color differences occurring in the post-processing step. If the paper is warm-white or cool-white has a much smaller impact on the realism of an arctic winter photograph than choosing appropriate white balance in post-processing. Those are fine-tuning problems.)
About here people get bored and think "it will never be accurate, so why care?" and just make something they find pleasing, not caring about realism at all. I'm fine with that, but some of us find this interesting and I prefer being able to talk about the subject without getting into meta discussions about what photography really should be about :). Sometimes I use the camera response creatively, long exposures in low light are very different from eye response for example, and sometimes I want a realistic rendering of the eye's experience at the scene, such as my arctic winter example. I think it becomes more rewarding to capture extra-ordinary light if it can be realistically reproduced.
In one aspect capturing extra-ordinary light on film was more rewarding than it is on digital. Film did not necessarily render it realistically, it have less potential to do so than digital, but the color conversion was the same so something extra-ordinary in real life became something extra-ordinary in the picture. With digital you just play around with contrast and saturation so everything looks extra-ordinary in the same way. You can be disciplined and limit your manipulation, but in extreme light I find that a challenge arise to find a realistic starting point.
Even in less extreme light I'd be glad if I could have a more realistic representation of colors, I've been shooting fall colors recently with my medium format system and I just did not manage to replicate a color response that seemed true to the scene. Either too brown or too green, couldn't find a proper mixture of yellow and greens. I'm not sure if it was my memory failing me or something else. In the end I chose something that looked pleasing, although I'd prefer the original if I could have replicated it. So I have become interested in what techniques there are to reproduce realistic colors. Due to the brain's "evening out" effects maybe the most true-to-the-original-scene impression is not given by the most accurate colors, there may be some psychovisual effects involved.
-
One problem is the brain's evening-out-effect, most clearly demonstrated with mixed artificial light. If you photograph that and want to have a realistic representation you will feel that you need multiple white balances, to bring the lights closer together as it was experienced on scene.
Not sure if you got my point. If you had a 13"x19" window in your darkened mountain cabin offering a view of the arctic valley below, would you feel a need to WB it? If you swapped this window for a 13"x19" print (with identical spectral characteristics), would you feel a need to WB it? What if you lit the interior of your cabin with candlelights and/or swapped the cabin with a city flat at summer time?
I think that the experience-based reliance upon WB concepts may be counter-productive in cases like this. I might be wrong.
-h
-
Not sure if you got my point. If you had a 13"x19" window in your darkened mountain cabin offering a view of the arctic valley below, would you feel a need to WB it? If you swapped this window for a 13"x19" print (with identical spectral characteristics), would you feel a need to WB it? What if you lit the interior of your cabin with candlelights and/or swapped the cabin with a city flat at summer time?
I think that the experience-based reliance upon WB concepts may be counter-productive in cases like this. I might be wrong.
If I look on reality through a window in a dark room there's no difference from standing outside. You'll still see that the arctic dusk light is blueish indeed. If the room is lit with ordinary room lights you'll experience the brain's "evening out" effect, you'll see that the outdoor light really is blue and the indoor is much much warmer but it does not really look strange. If you take an indoor-white-balanced photograph of the same you'll see the outdoor scene is much more extremely blue than it was to the eye, and if you adjust the WB so the outdoor look realistic the indoor looks wrong, so in that case you'll need dual white balance, and this is a problem architecture photographers often need to relate to.
Let's say we make a print with identical spectral characteristics, we would have to have it a backlit print to make it work. I think it would be experienced as looking out through the window. To make a regular print that works in ordinary display conditions you may need to make some adjustments, probably even it out a bit, ie maybe reduce the blueishness of the light slightly, but then we're in the fine-tuning phase.
-
If I look on reality through a window in a dark room there's no difference from standing outside. You'll still see that the arctic dusk light is blueish indeed. If the room is lit with ordinary room lights you'll experience the brain's "evening out" effect, you'll see that the outdoor light really is blue and the indoor is much much warmer but it does not really look strange. If you take an indoor-white-balanced photograph of the same you'll see the outdoor scene is much more extremely blue than it was to the eye, and if you adjust the WB so the outdoor look realistic the indoor looks wrong, so in that case you'll need dual white balance, and this is a problem architecture photographers often need to relate to.
Let's say we make a print with identical spectral characteristics, we would have to have it a backlit print to make it work. I think it would be experienced as looking out through the window. To make a regular print that works in ordinary display conditions you may need to make some adjustments, probably even it out a bit, ie maybe reduce the blueishness of the light slightly, but then we're in the fine-tuning phase.
If you make sure that the spectral characteristics of the print subject to whatever the lighting is inside the cabin matches (as closely as possible) the spectral content measured previously, should you need to do any tinkering? Should you need for the paper to be backlighted?
-h
-
Some philosophy:
Working with color and decide how to relate to it is an ongoing process for me. I know I don't want to end up like a photographer that thinks for a true artist anything is allowed and pull and push contrast and saturation to extreme levels, and on top of that say "I don't use photoshop much". I know many do, some very commercially successful, and good for them, but that's not how I want to develop my own photography.
I want to have some documentary aspect in my pictures, an honesty if you like. Even if it costs me a facebook like or two ;). I've put quite a large effort in how to use dynamic compression techniques for backlit scenes that need it and still retain a realistic look, and I think I've found quite good techniques for that (and even developed own software for that which I hope to be able to release soon). Realistic color rendering is an ongoing challenge though, and it's interesting to see that so little has been done. Is it because no-one wants it, because it's difficult or impossible to get meaningful results, or because no-one has had the time to do it yet... I'm not sure.
I don't believe in that healthy eyes see colors vastly different and thus any kind of measurement or matching would be meaningless as some imply. Current color models assume that we all see colors in a similar way. However different photographers with the same goal of realism still have different taste in how to adapt colors for a print, and different taste of how to fine-tune to make something look realistic. For example exactly how much should we reduce haze on that distant mountain, how much to compress highlights and adjust contrast. Some may think that the latitude of those post-processing changes is so large that some sort of "reference starting point" would be meaningless. I do believe that absolute accuracy is unnecessary for these reasons, but I also do believe that there could be better techniques than exist today.
-
If you make sure that the spectral characteristics of the print subject to whatever the lighting is inside the cabin matches (as closely as possible) the spectral content measured previously, should you need to do any tinkering? Should you need for the paper to be backlighted?
I think it becomes quite theoretical as I suspect that there are no techniques that are remotely close to replicate the spectral characteristics. But if you could and the walls in the cabin would not reflect any light, no you should not need to do any tinkering, it would look like looking out a window from a dark room, and the eye/brain would make the same adaptations as it would for the real scene. Ie I'm quite sure that it's 2D would not make any significant difference in color experience.
-
torger you seem very knowledgeable about the subject which makes me wonder why you posed the question in the first place?
Yes, this seem to be a thread moving into an area where one is searching for a solution in search of a problem.
-
torger you seem very knowledgeable about the subject which makes me wonder why you posed the question in the first place?
I've learnt a lot in a quite short period, and that may give you a bit of patchy knowledge, so I asked to see if there was any technique I did not know about and is missing out on. I also wanted to test if my assumption was true, ie that the current color models is not designed to solve these type of problems. And to see how other users are dealing with it. And the result seems to be that no there was no other technique, but the efficiency of manual WB live view on site matching is worth evaluating some more, and yes the assumption was true (color models are not designed to solve these type of problems), and most users deal with it by not thinking it is a problem.
I think the major disagreement we see in the thread is about if there actually is a problem or not, and in the extent there is if it's meaningful (or even possible) to solve or not. I thank everyone for their input.
-
Correct me if i am wrong ???
but I thought that if the light does not come from a black warm body- it does not have a complete spectrum of colours so in that light you cannot reflect all colours.
examples are TL-tubes and laserlight.
of course it depends on the definition of 'accurate colour' .
But it is clear that red laserlight shining on a blue subject will never be reflected as blue. So cannot be reproduced in a photograph as blue.
-
Correct me if i am wrong ???
but I thought that if the light does not come from a black warm body- it does not have a complete spectrum of colours so in that light you cannot reflect all colours.
Hi Pieter,
That's correct. Which begs the question, in what way does the arctic light differ from more regular daylight, which is seems to do. I'm a bit puzzled as to why the human eye, apparently, is not capable to compensate which thus creates a desire to also introduce a cold color cast in rendered images.
BTW, I am aware of the existence of blue ice (http://en.wikipedia.org/wiki/Blue_ice_(glacial)).
Cheers,
Bart
-
...Realistic color rendering is an ongoing challenge though, and it's interesting to see that so little has been done. Is it because no-one wants it, because it's difficult or impossible to get meaningful results, or because no-one has had the time to do it yet... I'm not sure.
As a former photo realistic styled painter, I can answer that with some authority with regard to emulating a 3D scene on a 2D surface and it is has a lot to do with mimicking the spectral reflectance "effect" we view in a real scene with our eyes bombarded with a huge surround of light compared to viewing within the reflective/transmissive environment of a flat, focused framed surface and what it does to our perception of the "real" scene on a display and print from an emotional aspect. There's a huge change to the perception many aren't and/or can't be aware of.
If we were to actually duplicate exactly what our eyes sees normally without the frame of a viewfinder, it looks rather dull and low contrast if you're REALLY honest with yourself on making that match. You can do it quite easily by creating a diorama scene of objects arranged together next to your display lit by a full spectrum light source such as your window or Solux lamp, photograph it in Raw and view it on your display. In my experience I can get it to look exactly as the diorama appears, but as for a desirable scene you'ld want to view for any length time? Not a chance. Pretty dull looking.
Can't do this with landscapes of course because you have to rely on memory of the scene which isn't exact at all. Pleasing is all one has left to render it as faithfully as possible and still maintain some emotional connection first established when tripping the shutter.
Could you post a sample image of a scene you think has "realistic" colors? That way we all can be on the same page on what you mean by this term which may not be what we consider as realistic at all.
-
I'm a bit puzzled as to why the human eye, apparently, is not capable to compensate which thus creates a desire to also introduce a cold color cast in rendered images.
The human eye (actually, not the eye but the brain it connects to) does compensate, even for Arctic light. Just imagine you are standing around photographing a Polar bear when the temperature is --40... Of course your brain will see everything as "cold"...
All jokes aside, the compensation actually is there. It just works a little different. And specifically a camera does not compensate. In a typical Arctic location there are fewer dark saturated colors. Water, tundra, rocks, and snow or ice are very common. They reflect the skylight better than more saturated colors found in more temperate regions. So either blue from a cloudless sky or gray clouds with blue filtered light will have more effect than would be so otherwise. Plus there is "Arctic Haze", or ice crystals in the air that filter light in a way not seen at warmer temperatures.
So the camera will see and record what actually is there. Lots of pastel skylight with a blue cast to everything.
And humans will see most landscapes as exactly that too. Not much compensation needed because that is what it should look like. But if we look at something that should be different, such as houses, people, cars... most of them 1) have more saturated colors, and 2) will look "normal" to us even in Arctic light.
-
Hi Bart,
I am pretty impressed that we can reconstruct several millions of colors from just three integrated quantities. I think it is quite natural that there is some ambiguity in that interpretation.
Best regards
Erik
Hi Pieter,
I'm a bit puzzled as to why the human eye, apparently, is not capable to compensate which thus creates a desire to also introduce a cold color cast in rendered images.
BTW, I am aware of the existence of blue ice (http://en.wikipedia.org/wiki/Blue_ice_(glacial)).
Cheers,
Bart
-
I thought that the core problem here was that the illumination is the pastell colored sky (or at least that is the brightest object in the scene, making everything else visible). Only when WB applied as usual, one tries to cancel the "non-whiteness" of that illuminant, thereby removing what is considered a desirable element?
It may well be that the higher-levels of perception in our mind is capable of holding several representations at the same time. Perhaps we can "white-balance" a scene (so as to make "red" objects appear "red") while simultaneously holding an internal reference to the overall/illuminant color temperature for reference?
-h
-
So the camera will see and record what actually is there. Lots of pastel skylight with a blue cast to everything.
I doubt this, it is a limited system with interpretations sort of baked in by the manufacturer.
-
I doubt this, it is a limited system with interpretations sort of baked in by the manufacturer.
If it were the camera you would see exactly the same thing in pictures from lower latitudes.
Oddly enough, even in BW the difference in Arctic light shows up! When I first started photographing the Arctic it was using Tri-X 35mm film with a Pentax Spotmatic in the area around the Bering Straits. It was just wierd! And what the film showed was almost exactly what it looked like too! And color slides didn't look much different! On a normal day in March there is dark water with white ice floating in it, snow covered land, and clouds that reflect either the darkness of the water or the whiteness of the land depending on visual perspective. No color.
-
Could you post a sample image of a scene you think has "realistic" colors? That way we all can be on the same page on what you mean by this term which may not be what we consider as realistic at all.
Thanks for your elaborate reply. I have a few examples earlier in this thread, the post with two pictures attached, one white-balanced for the snow and one fine-tuned after memory (with help of an fellow photographer too), both shot at the same occassion. The point there is that during those arctic light conditions it's not about fine-tuning within a small range, the light is so extreme that it can become hard to find what's right if there's been some time from shooting to processing.
Anyway, I may be mistaken for the typical beginner that thinks what the camera captures is what the eyes see and the rest is just manipulation. However I'm quite aware of the challenges, dynamic range is one of the obvious, ie no matter how hard I stare at my printed sunset I won't get blinded. We need to translate a huge dynamic range into a narrow one, and that means increasing contrast to sacrifice shadows and highlights, and tonemap.
The problem I'm looking at in this thread is hue (and saturation). I experience it difficult in some light conditions to be able to remember and reproduce a hue of the scene that is a realistic representation of the hue as experienced at the scene. Ie, was the colored band in the sky more towards red-pink or blue-magenta, and the color of the snow. Unlike dynamic range I find it less of a need to modify hue to make a good print. I know that you can modify hues and whiten clouds etc to improve the impression of contrast and saturation. I also know about that if you shoot a portrait with someone with a very saturated sweater some light can reflect into the face which you don't think about in reality but causes a tint in the photo, so there are exceptions.
However in the more extreme light conditions I've had in my examples (arctic dusk) it's not about these type of fine-tunings, it's about finding a starting point so you basically know what you're doing. To anyone that has not experienced these type of light conditions I can say it's pretty fantastic, it's so fantastic that I get a desire to be able to show how it really looked in an as realistic way as possible, adapting to the limitations of the medium. Being able to capture and reproduce a reasonably correct hue I think would be a good starting point.
-
To elaborate a little bit more with how I look upon "realism"; say if I have a large number of non-photagraphers with me when I make a shot, and then show them the print I made afterwards and ask them if they think it's an honest reproduction of the scene, if they think this is close to "how it actually looked" and most say yes, then I've succeeded.
There is of course no scientific formula of how you should translate an image capture to a print to succeed with this. You'll have to make a number of manipulations to make it look realistic, mainly contrast and tonemapping (which could be just a gradient, I'm not referring to grunge-looking HDR here :) ), and there's a lot of personal taste involved on which aspects you think is important to realism or not. For example you may think that a sense of strong contrast is more important than hues, so you whiten your clouds etc to strengthen the sense of contrast. I on the other hand think that hue is a more important aspect and would like to focus on getting those close to the at-scene experience.
What type of modifications you might need to do to the hues due to different viewing conditions I've not gotten into so much yet, but it may not be unlikely that for that arctic light you may need to dampen the blueishness a bit to not make it too unbelievable.
In any case I would appreciate having some sort of tools to be able to have a better baseline on hues when I start to work than exists today.
(I've noted that there seems to be a difference between say american and european audience. The realism/documentary aspect is more popular here in europe, I doubt photographers like Peter Lik or Rodney Lough would be as successful with their printing style on an european market. Adapting a more realistic style does not mean that you totally drop your artistic interpretation and just document, the artistry is in chosen composition, which images you choose to show, and of course you probably still do some dodge and burn and similar to focus your scene in the direction you want and adapt to the printing medium.)
-
I thought that the core problem here was that the illumination is the pastell colored sky (or at least that is the brightest object in the scene, making everything else visible). Only when WB applied as usual, one tries to cancel the "non-whiteness" of that illuminant, thereby removing what is considered a desirable element?
It may well be that the higher-levels of perception in our mind is capable of holding several representations at the same time. Perhaps we can "white-balance" a scene (so as to make "red" objects appear "red") while simultaneously holding an internal reference to the overall/illuminant color temperature for reference?
Yes, I can guarantee that on my posted examples that the pastell color is so strong that a normal WB technique will render a much less realistic result than even if you just set the WB to daylight. If someone thinks I'm talking rubbish on that point it's not much more to discuss, then I'd just like to invite them to come and see when winter comes ;).
My posted examples are relatively "simple", as the light in them is fairly uniform. However, when the sun has not yet set you have both skylight in the shadows and direct sunlight on some parts of the image, those can be extremely difficult to get right. I think what happens there is similar to the effect with multiple artificial light sources, ie the illuminants are so extreme and separated that you can't do with one white balance.
I've attached one such example, I'm actually not entirerly happy with the rendering. More of a snapshot style of image but presented an interesting rendering challenge (both dynamic range and color) so I worked quite long time with it but did not succeed getting to a point which felt 100% right, but it's sort of close I think. Here I've employed a dual white balance, if I only had one white balance and balanced it to get the sun covered mountain yellow-red the snow got greenish, but that is actually more blueish in the shadow. There's also a gradient to fit the bright mountain without darkening the shadowed landscape too much.
I'm suspecting that in this example due to the dynamic range and extreme spectrums there's so much psychovisual things going on that it's pretty impossible to model. If I would try the "manual wb match with live view on site" white balance trick I'd have to make one reference shot for the shadowed landscape and one for the mountain. I hope to make some more photography this winter of this kind as I find this landscape very beautiful and also from a technical perspective very interesting and challenging.
I'm suspecting that someone that has not seen these kind of winter scenes would say "hey, there's a blue cast in your image" and they'd think a white-balanced for the snow would be closer to reality. I've tried that rendering too, but it's so clearly not right for someone that's been on site that I chose this rendering instead, which is the most realistic interpretation of the scene I could make with my current skill and techniques.
-
As an example in a recent discussion about RAW vs jpeg I posted a jpeg rendered by a digital camera:
(https://dl.dropboxusercontent.com/u/19059944/chania9b)
...then I posted an image that I edited in ACR on my Macbook Retina while watching at the scene (I have perfect score in FM100 hue test, and practiced oil painting while I was a student at Academy of Fine Arts):
(https://dl.dropboxusercontent.com/u/19059944/chania9)
Ironically, everyone said that my image doesn't look realistic and that the jpeg has more natural colour rendering :D
Conclusion is that people don't really want realistic colours - which is fine, because cameras are completely unable to render them...
-
Conclusion is that people don't really want realistic colours - which is fine, because cameras are completely unable to render them...
That is a terrific example! But I would differ slightly with what it demonstrates. It isn't so much that "people don't want realistic colours", because that is not a picture of the colors! It's a little bit of what Ansel Adams meant when he said "There is nothing worse than a sharp image of a fuzzy concept." What Pablo Picasso meant with "Art is the elimination of the unnecessary." and relates to that old saw about shooting in BW makes a photograph of a person's soul while color takes a picture of the color of their clothes.
To be a little more academic about it, Rudolf Arnheim (1904-2007), in "Film as Art" said,
"This discovery of the gestalt school fitted the notion that the
work of art, too, is not simply an imitation or selective
duplication of reality but a translation of observed
characteristics into the forms of a given medium"
The version that attempts to duplicate the "accurate" colors is wonderful if and only if you have a need to record the colors that existed. But that isn't the same as making an appealing photograph of the pool, the building, their relationship, or whatever else it might be that is a more useful photograph as a work of art.
If the building's relationship to the pool is the object, then the sky in the first one which is much less of a distraction, is a better choice. So might be the darker green of the trees in the background. But the second image makes the building more dominant, though the greens in the foreground could probably be reduced in dominance too.
It's not a matter of what people want, it's a matter of what the photographer wants the image to convey to the viewer! The image is a photograph (a form of communications, an art form) and is not reality. Editing needs to consider what is being communicated to the viewer, not what the scene "really" was when you saw it.
-
I agree about that you may edit your colors to convey a message, although I think one can do most by saturation and luminance, not hue. I'm also not going to stop shooting long exposures in low light or other cases when the camera can see things impossible with the eye. However, if I could build a post-processing style where I have a firm connection to the eye's experience at the scene, and I would know how much I deviate from that I'd be glad. I want to be subtle, and to be that I need to have some sort of "neutral" starting point which I think sometimes can be hard to find the way I work today.
All too often the feeling that I handle my colors pretty randomly with lots of guessing, and post-processing one day may be different from another day depending on my temper. I guess I'm in the process of trying to find some more structured method to work, and connecting to the color experience at the scene is one of the things.
Sometimes I get the case when I really liked what I saw in terms of color, ie I don't feel that there should be any need to play around with the hues, and try to adjust in post to get the right look as I remembered it, but fail to do so. In that case it would have been nice if there was a way to recreate it (but there isn't...).
-
The version that attempts to duplicate the "accurate" colors is wonderful if and only if you have a need to record the colors that existed.
And where it existed. Color, is a perceptual property. So if you can't see it it's not a color. Color is not a particular wavelength of light. We define colors based on perceptual experiments. Excitation of photoreceptors followed by retinal processing and ending in the visual cortex, this is stuff happening in our head and that's the colors that existed. Cameras operate on a different level with respect to 'color'.
-
And where it existed. Color, is a perceptual property. So if you can't see it it's not a color. Color is not a particular wavelength of light. We define colors based on perceptual experiments. Excitation of photoreceptors followed by retinal processing and ending in the visual cortex, this is stuff happening in our head and that's the colors that existed. Cameras operate on a different level with respect to 'color'.
This is exactly correct.
Tony Jay
-
And where it existed. Color, is a perceptual property. So if you can't see it it's not a color. Color is not a particular wavelength of light. We define colors based on perceptual experiments. Excitation of photoreceptors followed by retinal processing and ending in the visual cortex, this is stuff happening in our head and that's the colors that existed. Cameras operate on a different level with respect to 'color'.
Even though the word "color" is used to describe a perceptual thing, the world that enable us to sense colors is a very physical thing. Recreate all of the physical stimuli leading to some perceptual response, and chances are that you are (as close as you will ever get) to recreating that perceptual response once more.
If you have this pastel colored scene referenced by the OP, I don't get why color philosophy is needed to make it realistic. Rather, it seems to me that the standard practice of "white balancing away the light" is the problem, something which there should be pragmatic cures for?
-h
-
> Rather, it seems to me that the standard practice of "white balancing away the light" is the problem, something which there should be pragmatic cures for?
But obviously, there are not! I think the reason is that the standard practice is tailored to something like catalogue photography and art reproduction, where "White balancing away the light" is indeed needed - not to landscape shooting, where one wants to picture that same light and its reflections.
Good light and true color! - Hening
-
If you have this pastel colored scene referenced by the OP, I don't get why color philosophy is needed to make it realistic. Rather, it seems to me that the standard practice of "white balancing away the light" is the problem, something which there should be pragmatic cures for?
Philosophy aside, you'd need some multispectral camera and intelligent colour appearance model as a "pragmatic cure" in the above mentioned situation.
-
Even though the word "color" is used to describe a perceptual thing, the world that enable us to sense colors is a very physical thing.
Yes to some degree. We can measure that color and by doing so, we can have pretty good way to discuss inaccuracy using dE. We're talking about single colors, not colors in context. For calibrating and profiling device behavior, this model, colorimetry works well.
-
Yes to some degree. We can measure that color and by doing so, we can have pretty good way to discuss inaccuracy using dE. We're talking about single colors, not colors in context. For calibrating and profiling device behavior, this model, colorimetry works well.
So, for recreating the effect of "looking through a mountain lodge window"*) onto the scene... Calibrating and profiling the camera and display/print method would be sufficient (of course, you would be limited to the gamut/DR/... of the imaging chain)? Is this similar to using a standard observer as a reference, and minimising the error (given some constraints/regularization) that reproducing the scene using the current camera/display/print would give?
Does the display have to be self-illuminant (e.g. LCD) for this to work, or is it possible to use a (profiled) print along with a profile/guesstimate of the room lighting? (Constrained by the radiated power/reflectance available at each wavelength; a yellow-tinted candle lit interior might not be able to illuminate a convincing simulation of a blue-cast snowy landscape unless that landscape was close to darkness)
-h
*)If my window references seems odd: I just want to separate those errors that stems from filling the human FOV with one consistent scene, vs filling part of our view with one scene, and the rest with a (potentially highly) different scene. A mountain logde window seems like a real-world reference that we can relate better to than talking about "5 degree standard observers" or something like it.
-
Philosophy aside, you'd need some multispectral camera and intelligent colour appearance model as a "pragmatic cure" in the above mentioned situation.
Why is that?
A crude engineering approach might be:
1. Obtain a profile of the camera (wavelength vs sensitivity for each primary)
2. Obtain a profile for you display (energy vs wavelength for each primary)
3. For a given camera raw file, apply a linear/nonlinear transform that minimize the squared error within [390 700]nm (make each output pixel the closest possible spectrum to that which is known about the scene)
A less crude approach might be to minimize the error within bands corresponding to "red", "green" and "blue".
Having a robust multispectral camera would probably make things easier, but I don't see that it is critical unless the scene contains some "irregular" and "hairy" spectra. Are the OPs image likely to be irregular, or are they more probably smooth and easily characterised?
-h
-
Problem is that "images acquired with RGB cameras contains on the contrary a systematic color error, since no basic linear dependency can be found between the spectral sensitivity of RGB cameras and the spectral sensitivity of human observers":
http://www.lfb.rwth-aachen.de/en/research/basic-research/multispectral/
But the major problem is that the observer model is "wrong" - it works for D50 illuminated colours on 20% grey background in bright surround. The different colour context, viewing conditions, brightness, chromatic adaptation, and countless other factors make that it fails:
http://www.cis.rit.edu/fairchild/PDFs/AppearanceLec.pdf
When we take all these factors into account it becomes obvious, that there's not event the slightest chance to get "right" colours on OP's image - due to limitations of the camera sensor and RAW developing methods.
-
Problem is that "images acquired with RGB cameras contains on the contrary a systematic color error, since no basic linear dependency can be found between the spectral sensitivity of RGB cameras and the spectral sensitivity of human observers":
http://www.lfb.rwth-aachen.de/en/research/basic-research/multispectral/
That sentence seems odd to me. Is there _no_ linear dependency? I would assume that there is one, but that it does not allow for a complete transformation (i.e. there might be an "optimal" 3x3 matrix, but you would still have errors). I am an engineer, errors are a fact of life. The question is how much of a problem those errors are.
But the major problem is that the observer model is "wrong" - it works for D50 illuminated colours on 20% grey background in bright surround. The different colour context, viewing conditions, brightness, chromatic adaptation, and countless other factors make that it fails:
http://www.cis.rit.edu/fairchild/PDFs/AppearanceLec.pdf
If I want my framed image/LCD to look as if it was a window peeping into a landscape, most of those problems would go away, would they not?
I would think that the low-level spectral sensitivity of my sight is relatively stable (cones and rods). High-level processing is probably highly context-dependant.
When we take all these factors into account it becomes obvious, that there's not event the slightest chance to get "right" colours on OP's image - due to limitations of the camera sensor and RAW developing methods.
So there is not the slightest chance that results for the OP can be improved by a mere change of software algorithms/rendering goals?
It would be neat to have a Lightroom slider that went from "neutral illumination (relative colours)" to "original illumination (absolute colours)".
-h
-
I want my framed image/LCD to look as if it was a window peeping into a landscape, most of those problems would go away, would they not?
But how? The sensor introduces some "errors", and then the colours are interpreted as D50 2 degree stimuli on 20% grey in bright surround which is 100% not the case of OP image.
To capture "right" colours we would need a camera with a sensor that matches XYZ curves better (and maybe a spectroradiometer on top of it), and a RAW developer that takes into account viewing conditions, colour context, scene brightness and so on, and then renders properly interpreted colours into the editing space. Or maybe something even more complicated.
It would be neat to have a Lightroom slider that went from "neutral illumination (relative colours)" to "original illumination (absolute colours)".
Lightroom doesn't know what "original illumination" really was, that's the part of the problem.
-
So, for recreating the effect of "looking through a mountain lodge window"*) onto the scene... Calibrating and profiling the camera and display/print method would be sufficient (of course, you would be limited to the gamut/DR/... of the imaging chain)?
I don't know what calibrating and profiling the camera implies but rarely are the desired or obtained results colorimetry correct and may not match something else.
If you want to know the behavior of an Epson printer today and in a year, with the current tools, very useful and doable. You print out a target and measure it today and in a year, you do so on a device that calibrates to a reference and has good repeatability. You define how many patches and where in color space to sample the output. You can plot on a daily dE2000 behavior of your printer, compare just the paper (media dE), understand dry down behavior etc. None of this has anything to do with color appearance, or Colors in context.
-
@czornyj
>The sensor introduces some "errors", and then the colours are interpreted as D50 2 degree stimuli on 20% grey in bright surround which is 100% not the case of OP image.
And yet, the OP, Torger, has managed to re-create colors that are very very close (image in post #55) or at least very trustworthy and naturally looking, by memory. Now if he would eliminate the memory factor by comparing the actual view with a stepped scale of camera recorded white balances at shooting time, he might come even closer (if that is at all possible), couldn't he?
-
But how? The sensor introduces some "errors", and then the colours are interpreted as D50 2 degree stimuli on 20% grey in bright surround which is 100% not the case of OP image.
The figure below seems to depict the camera raw file signal vs scene spectrum:
(http://www.astrosurf.com/buil/50d/canon_40d_50d.png)
The figure below seems to depict LCD (using "white" and "rgb" LED backlight) spectral intensity:
(http://powerelectronics.com/site-files/powerelectronics.com/files/archive/powerelectronics.com/power_management/led_drivers/Fig-2-white-LED-vs-RGB-LED-spectrum.jpg)
Using information similar to the two figures (for a given camera and display), I guess that one might try to recreate (what is known about) the scene spectrum with some error using the freedom in the display. The engineering approach might be minimizing the squared error. I assume that something similar can be done for the 10 or so inks found in photo ink printers, except you will have to measure the spectrum of the combined ink+paper+illumination.
I don't think that D50, 2 degree stimuli, bright surround etc comes into play for this example as formulated by me.
Perceptual stuff will probably be relevant once you try to make the error criterion better suited for the goal of making images that is to be viewed by humans (some spectral errors might be more critical than others).
Still, if you could measure the spectrum of every pixel within 400-800nm with a very small squared error, and recreate it with a very small squared error (using multi-spectral means), a human would probably think that the scene was "accurately" recreated if framing, size etc was kept constant. Doing the same using a (less accurate) 3-channel colour aparatus is conceptually the same, only with larger error.
-h
-
And yet, the OP, Torger, has managed to re-create colors that are very very close (image in post #55) or at least very trustworthy and naturally looking, by memory. Now if he would eliminate the memory factor by comparing the actual view with a stepped scale of camera recorded white balances at shooting time, he might come even closer (if that is at all possible), couldn't he?
Yes, to some degree, like I did myself in case of sample posted here:
http://www.luminous-landscape.com/forum/index.php?topic=82738.msg668928#msg668928
The problem is the difference of scene contrast, display gamut, and the fact, that you can spoil some colours by editing others, but the overall impression was much closer to the look of the scene. I suppose at this moment direct comparison is the simplest solution to get the "right" colours.
-
Using information similar to the two figures (for a given camera and display), I guess that one might try to recreate (what is known about) the scene spectrum with some error using the freedom in the display. The engineering approach might be minimizing the squared error. I assume that something similar can be done for the 10 or so inks found in photo ink printers, except you will have to measure the spectrum of the combined ink+paper+illumination.
Of course we can't. We have completely no idea what was the spectrum captured by sensels of the camera. And even if we knew it, we couldn't recreate it with the display pixels.
I don't think that D50, 2 degree stimuli, bright surround etc comes into play for this example as formulated by me.
Of course it does - we correct the difference between the sensor's spectral curves and XYZ curves of standard D50 2 degree observer using camera profiles, then we display colours that produce metameric match on a monitor. The colours on the display have completely nothing to do with the original spectra, which BTW is unknown.
-
Of course we can't. We have completely no idea what was the spectrum captured by sensels of the camera. And even if we knew it, we couldn't recreate it with the display pixels.
If one knows the spectral sensitivity of each color channel of the camera, one _does_ have an idea what the spectrum was. The measurement may not be very "accurate", but that was never my claim. Claiming that you have "completely no idea what was the spectrum" is clearly wrong.
Of course it does - we...
Who are "we"? Me? you? Adobe? I was trying to be very clear about my suggestion in my post, yet you seem to put all kinds of assumptions in-between my lines. Please don't do that. Please re-read my post and interpret it literally. If I am unclear or you are confused, please ask instead of assuming.
...correct the difference between the sensor's spectral curves and XYZ curves of standard D50 2 degree observer using camera profiles, then we display colours that produce metameric match on a monitor. The colours on the display have completely nothing to do with the original spectra, which BTW is unknown.
Who mentioned XYZ? Who mentioned camera profiles?
-h
-
I'm just trying to explain how it works.
Please, rethink it once again - how could we predict the original spectra of the scene, when countless different spectra can induce the same RGB signal?
How could we reproduce the original spectra using display that has a matrix with only three colour filters?
-
I'm just trying to explain how it works.
And I was offering suggestions to how the OPs problem might be improved.
I am not a color scientist, so my suggestions might be naive or impractical. If so, I'd like to understand why it is so.
Please, rethink it once again - how could we predict the original spectra of the scene, when countless different spectra can induce the same RGB signal?
How does a carpenter measure the length of a wall when the measure does not have infinite precision? He take the best reading that he can, make the best out of it, and (ideally) keeps a reference of the uncertainty.
How could we reproduce the original spectra using display that has a matrix with only three colour filters?
You use what you have to the best of your abilities.
A "rgb" camera or display may be seen as a somewhat irregular 3-channel filter bank (spectrally). By adjusting the gain of the 3 channels, you can get some level of approximation to some spectrum. If you had a 256-channel uniform filterbank, the _problem_ would not be fundamentally different (in my view), but the _precision_ of the reconstruction would be a lot better.
Think about it - even with e.g. 256 uniform spectral channels, there will still be limited spectral precision. You will still have ambiguity in that two different physical scenes can produce the same sampled signal (and thus impossible to separate or recreate accurately).
I am not convinced that the problem of the OP is really that the spectra are "hard" to record accurately (i.e. "hairy" spikes that can only be recorded using multi-spectral means, or at least color sensing aparatus that is a very good approximation to human visual perception). Rather, I have a feeling that the colors are (in a way) accurately recorded by his camera and representable by his display/printer. But somewhere along the line, our "best practice" color processing fails his task. By e.g. attempting to remove absolute WB. If "absolute scene white-point" could be established (by camera profile, or pointing a physical "white picker" towards the pastel sky), perhaps simple changes to color processing could attempt to make this the reference point (instead of "neutral white")?
-h
-
And I was offering suggestions to how the OPs problem might be improved.
Not using colorimetry or the current behavior and options we have to capture images in the field. This was a problem in search of a solution (actual solution based on today's tools and techniques, move the sliders and make the images look as you prefer). Now we're stretching into science fiction in terms of what might be possible in an attempt to solve the problem? Czornyj has explained the issues well and asked the right questions (how could we predict the original spectra of the scene, when countless different spectra can induce the same RGB signal?). IF we had the spectral data captured with the image, IF we had the spectral sensitivities of the sensor, we'd be in a different possible position here. We have neither.
How does a carpenter measure the length of a wall when the measure does not have infinite precision?
Doesn't wash. We know what data we need just to start this process. You should be asking, How does a carpenter measure the length of a wall when the measurement is only provided in a weight, not a distance. Answer, he can't.
-
Not using colorimetry or the current behavior and options we have to capture images in the field. This was a problem in search of a solution (actual solution based on today's tools and techniques, move the sliders and make the images look as you prefer).
The OP clearly did not like the solutions of "move sliders until you are happy". Taken to the extreme, this is an argument against all color management. Just keep printing and pushing sliders until you are happy? Not a very predictable or satisfactory process.
Czornyj has explained the issues well and asked the right questions
Czornyj has explained how stuff works today. He did not explain why it has to be so, or (as far as I have seen) suggest a solution.
(how could we predict the original spectra of the scene, when countless different spectra can induce the same RGB signal?).
I am surprised that I am so poor at delivering this point. Are you disputing that 3 parameters can be used to describe a spectrum (with some precision)?
IF we had the spectral data captured with the image, IF we had the spectral sensitivities of the sensor, we'd be in a different possible position here. We have neither. Doesn't wash. We know what data we need just to start this process. You should be asking, How does a carpenter measure the length of a wall when the measurement is only provided in a weight, not a distance. Answer, he can't.
A weight does not relate to a tape measure like 256 channels does to 3 channels. Weights and tape measures measures two different things. 3-ch and 256-ch cameras measure the same thing with different precision.
It seems that spectral sensitivites of camera sensors are available. If your Antarctica trip was ruined due to poor colors, I am sure that you can afford to have some lab measure the sensitivity of your camera with great precision. I agree that true spectral data of a point of a given scene is more of a long shot.
-h
-
The OP clearly did not like the solutions of "move sliders until you are happy".
I'm aware of that. So his options are don't move the sliders and get what he gets, set the camera to capture a JPEG and get what he gets, hope there is some science fiction like product or device that gives him an exact replica of his memory. Might I point out that I hate flying and would prefer a transporter device but as yet, that's all science fiction. I fly, he moves sliders (or doesn't).
Taken to the extreme, this is an argument against all color management.
Taken to the extreme, this is an argument agains taking photo's. You can take anything to the extreme including speculating about science fiction but today, that's not a useful process.
Just keep printing and pushing sliders until you are happy?
Or any similar control to edit your images, otherwise, accept what the automatic system provides and move on.
Czornyj has explained how stuff works today.
Indeed, and unless you want to speculate and move into the realm of science fiction, a subject that's probably more appropriate in another forum on a different web site, why go there?
He did not explain why it has to be so, or (as far as I have seen) suggest a solution.
He and I did. Just get a tiny and inexpensive Spectrophotometer built into the camera that triggers each time the shutter does, and of course write that EXIF data to the raw, get the spectral sensitively of the chip. You're pretty much there at least in terms of defining the color of the scene and the capture device. There's more work to be done, we still need a good color appearance model because again, current technology doesn't treat colors in context like we do.
It seems that spectral sensitivites of camera sensors are available.
Where (and by whom)? And how will you account for the capture colorimetry portion? It's not impossible, it's massively expensive and complex. But then I'm told a transporter isn't impossible either, but rather:
In 1993 an international group of six scientists, including IBM Fellow Charles H. Bennett, confirmed the intuitions of the majority of science fiction writers by showing that perfect teleportation is indeed possible in principle, but only if the original is destroyed
-
Hi,
I dont really now. The samples the OP posted makes me consider mixed light. Part of the image is illuminated by skylight and part by orange shifted sunlight. The humidity of the air affects the amount of blue removed from the sunlight. So we have say 20000K skylight and 3000K sunlight illuminating different parts of the image and this dual illumination is part of the visual impact of the image.
I guess that modern DSLRs are probably handling lighting like this well, basically choosing a white balance with both satisfactory reds and blues. Getting the right WB is the first order solution after that we may need to fix some second order effects.
Best regards
Erik
I'm aware of that. So his options are don't move the sliders and get what he gets, set the camera to capture a JPEG and get what he gets, hope there is some science fiction like product or device that gives him an exact replica of his memory. Might I point out that I hate flying and would prefer a transporter device but as yet, that's all science fiction. I fly, he moves sliders (or doesn't). Taken to the extreme, this is an argument agains taking photo's. You can take anything to the extreme including speculating about science fiction but today, that's not a useful process.
Or any similar control to edit your images, otherwise, accept what the automatic system provides and move on.
Indeed, and unless you want to speculate and move into the realm of science fiction, a subject that's probably more appropriate in another forum on a different web site, why go there? He and I did. Just get a tiny and inexpensive Spectrophotometer built into the camera that triggers each time the shutter does, and of course write that EXIF data to the raw, get the spectral sensitively of the chip. You're pretty much there at least in terms of defining the color of the scene and the capture device. There's more work to be done, we still need a good color appearance model because again, current technology doesn't treat colors in context like we do. Where (and by whom)? And how will you account for the capture colorimetry portion? It's not impossible, it's massively expensive and complex. But then I'm told a transporter isn't impossible either, but rather:
-
I'm aware of that. So his options are don't move the sliders and get what he gets, set the camera to capture a JPEG and get what he gets, hope there is some science fiction like product or device that gives him an exact replica of his memory. Might I point out that I hate flying and would prefer a transporter device but as yet, that's all science fiction.
If one cannot (or should not) discuss why plains are like they are, then how can planes ever improve? Rather than try to ridicule my posts by calling it "science fiction", why don't you try to use your color expertise to explain why you think it is the wrong direction?
You still havent explained what the _fundamental_ difference of a 3ch color device vs a 256ch color device is except the obvious (number of channels). Why is it that a 3ch device cannot "predict the spectrum", while a 256ch device can? If so, where is the limit? Can a 4ch device "predict the spectrum"? or a 23 ch device?
I fly, he moves sliders (or doesn't). Taken to the extreme, this is an argument agains taking photo's. You can take anything to the extreme including speculating about science fiction but today, that's not a useful process.
Or any similar control to edit your images, otherwise, accept what the automatic system provides and move on.
Taking things to the extreme can be an efficient method to make authors of poorly founded claims think things over.
Indeed, and unless you want to speculate and move into the realm of science fiction, a subject that's probably more appropriate in another forum on a different web site, why go there? He and I did. Just get a tiny and inexpensive Spectrophotometer built into the camera that triggers each time the shutter does, and of course write that EXIF data to the raw, get the spectral sensitively of the chip. You're pretty much there at least in terms of defining the color of the scene and the capture device. There's more work to be done, we still need a good color appearance model because again, current technology doesn't treat colors in context like we do. Where (and by whom)? And how will you account for the capture colorimetry portion? It's not impossible, it's massively expensive and complex. But then I'm told a transporter isn't impossible either, but rather:
Your continued attempts at ridicule makes for a less interesting discussion and makes you look less competent than I think you are. Can you please stop?
I don't get why we need a color appearance model if the goal is to reproduce an image as if one was looking at the scene through a window. Can you explain this?
I think that part of the confusion stems from unwritten assumptions about the desirability of making a landscape image appear as if one is looking through a window. Do you think that this can be desirable, or do you think that all images should be "perceptually corrected"?
-h
-
If one cannot (or should not) discuss why plains are like they are, then how can planes ever improve? Rather than try to ridicule my posts by calling it "science fiction", why don't you try to use your color expertise to explain why you think it is the wrong direction?
I have, several times. Did you not note the colors in context discussion? Or that we need spectral data of the scene illuminant and the sensor to even approach the data needs for some kind of perceptual color modeling? Or that color is something that happens between your ears, using that large organ called your brain. In the context of this post, color is not a particular wavelength of light. It isn't a set of numbers that defines a single color pixel.
You still havent explained what the _fundamental_ difference of a 3ch color device vs a 256ch color device is except the obvious (number of channels).
You are asking the wrong question. The question should be, do we have spectral data? With the camera, we do not. So you might have missed my point about needing a built in Spectrophotometer in our cameras and recording actual spectral data of the illuminant within each capture.
Your continued attempts at ridicule makes for a less interesting discussion and makes you look less competent than I think you are.
With all due respect, please take my comment in context to this: you obviously mistake me as someone who gives a s*&t.
I don't get why we need a color appearance model if the goal is to reproduce an image as if one was looking at the scene through a window. Can you explain this?
I told you for at least the 2nd time above. We're dealing with numbers of solid colors without reference to other colors, the current technology has a slew of perceptual flaws of which some are working to overcome with better color appearance models. The phenomena that happens inside your brain and not thousands of solid colors that when viewed on a display resemble an image. Not what you recall you thought you saw at the scene. I posted a URL on the extreme difference between scene and output referred imagery. Start here (http://www.michaelbach.de/ot/col_mix/index.html and then http://en.wikipedia.org/wiki/Color_model) and try to understand how complex human perception is to model considering the slew of conditions where our vision is fooled.
Do you think that this can be desirable, or do you think that all images should be "perceptually corrected"?
What the OP wants IS desirable but unfortunately about as likely as transporting humans at this point in time. I have no idea what you mean by "perceptually corrected" but if that is supposed to mean we have to manually adjust images based on the current tools and capture technology to get a pleasing image, an image that isn't anything close to being colorimetrically correct, the answer is yes (and I said that already, probably more than once here).
-
But if I compare the camera screen to the real scene at shooting time, are not the "things that happen inside my brain" the same for both the scene and the screen ? (omitting the bias of the jpeg for now). Could this trick not move the image a bit from "what I recall I thought I saw at the scene" to what I really saw?
I note that Czornyj, who seems to be your peer in knowledge, and with whom you agree in describing the problem, arrives at a conclusion quite different from yours: "The problem is the difference of scene contrast, display gamut, and the fact, that you can spoil some colours by editing others, but the overall impression was much closer to the look of the scene. I suppose at this moment direct comparison is the simplest solution to get the "right" colours." (Post #73).
That isn't QUITE the same as "pull the sliders until you're happy", is it well?
-
That isn't QUITE the same as "pull the sliders until you're happy", is it well?
I pulled the sliders until I get similar look on the screen, that's the only difference. I did it just to illustrate the fact, that the camera renders the "wrong" colours in other discussion. The result was that I was criticised for achieving artificial effect by people who didn't even saw the original scene, but stated that jpeg rendered by the camera looked more realistic :D
-
But if I compare the camera screen to the real scene at shooting time, are not the "things that happen inside my brain" the same for both the scene and the screen ? (omitting the bias of the jpeg for now).
That bias is huge, and again, it's output referred. It's squashed so to speak, to fit the range of that screen (squashed in terms of dynamic range, gamut etc). Consider the illuminant of the LCD and the role it has on what you're viewing at that moment.
I note that Czornyj, who seems to be your peer in knowledge, and with whom you agree in describing the problem, arrives at a conclusion quite different from yours: "The problem is the difference of scene contrast, display gamut, and the fact, that you can spoil some colours by editing others, but the overall impression was much closer to the look of the scene.
That's indeed a big part of the 'problem' along with the other issues I brought up. Our conclusions are the same and we're talking about different problems that make this so difficult.
-
I can not see that your conclusion is the same as Czornyj's. You reject everything if it is not colorimetry; he accepts a partial improvement based on visual comparison. So do I. I think that a visual comparison done at shooting time has a chance to come closer to what I saw than one done by memory - even with all the biases. That's all. That's what I can do with todays tools, and I find it a little more satisfactory than shear arbitrariness.
-
I can not see that your conclusion is the same as Czornyj's. You reject everything if it is not colorimetry; he accepts a partial improvement based on visual comparison.
I thought my writing here and elsewhere, other posts were clear. Let me try again. I reject the term accurate and color without colorimetry. Without colorimetry we have no basis for the use of the term accurate. I accept pleasing or desired color that isn't colorimetry correct. By it's very nature, pleasing output referred color is rarely if ever colorimetrically correct. If you look at the LCD (which is a canned rendering), dismiss the illuminant of that device but see it matches what you view next to it, that's fine. Got no problem with that. Calling it accurate is simply not true and can't be proven with any metric. It's subjective. I have no issue with subjective and pleasing color. It IS what nearly everyone here is producing. With differing degrees of work.
I think that a visual comparison done at shooting time has a chance to come closer to what I saw than one done by memory - even with all the biases. That's all. That's what I can do with todays tools, and I find it a little more satisfactory than shear arbitrariness.
And that's fine! But it isn't accurate color. It's what you want, expect, accept, pleased by. Unless you can measure a number of items and 'do the math', calling it accurate isn't an accurate statement. Is that clearer picture of my POV?
-
Hi
May I chip in here. I have been following this thread and the Op question is "Is accurate color possible in non-standard light'
(Or do you mean 'accurate colour appearance'? There is a difference but thats a whole new can of worms)
Almost is the answer, but not practical with a single shot capture and current white balance algorithms.
I think the main problem here is white balancing the scene.
Humans can view a scene with different light sources and are constantly white balancing as we look around (remember most of our colour vision is in the 2˙ field) and discounting the illuminant, so in effect we perceive a sort of uniform white balance and a perceived colour constancy across the scene.
A camera (or post software) must make a single white balance prediction across the frame, not a pixel by pixel choice.
White point estimation is, using a single shot RGB capture, an under-constrained problem.
To calculate the illumination colour we need the surface spectral reflectance properties and the RGB device responses.
We only have the device RGB values.
There are several white balance algorithms of varying complexity, Grey World , Max RGB, Hybrid Retinex etc. They can produce good results but can fail drastically if the range of scene colours is limited, as in OP scene.
If you are shooting RAW WB is done in post but will rely on 'memory of scene" if done off site or if done in the field would require well calibrated monitors with a wide a gamut as possible under the appropriate viewing conditions, preferable not a laptop! But again you are white balancing across whole scene, so after chromatic adaptation to output space White Point (D65, or D50), some parts may look correct, others will be out.
Multi Chanel capture will be better at WB that a 3 channel. A 256 device would be like a spectrophotometer! Good but impractical.
As mentioned by Andrew, a small spectro in the camera at time of capture would help but the problem of multi scene illumination would still arise.
Introducing a colour appearance model would go some way to achieving the OP goal, but this again requires robust White Point estimation to calculate scene appearance prior to rendering. (See attached)
One would really need pixel by pixel WB and Appearance modelling, so come on Nikon, Canon Adobe, get this sorted..
Iain
-
With all due respect, please take my comment in context to this: you obviously mistake me as someone who gives a s*&t.
With all due respect, I believed that you had something to contribute, other than cursing and ridicule. Have a good day, sir.
-h
-
I think the main problem here is white balancing the scene.
Humans can view a scene with different light sources and are constantly white balancing as we look around (remember most of our colour vision is in the 2˙ field) and discounting the illuminant, so in effect we perceive a sort of uniform white balance and a perceived colour constancy across the scene.
I agree that WB seems like a problem.
If you can bear with my analogies: a 13"x19" window on the wall of my cabin might cover a large or small part of our vision (depending on how close you view it). It may also have a radically different WB/spectral properties from that of the cabin interior. So that vista may appear "blue cast". Or any other tint. My thesis is that this "color cast" is acceptable, and perhaps desirable in this particular setting. Now swap that window with an LCD screen or a passively illuminated print. Do we still want that "color cast"? Again, my thesis is "yes", at least for some images.
How do we recreate the appearances of looking through that window (or, stricter but sufficient: how do we recreate the physics of the light that was flowing through the window, that triggered our perceptual response), and how does this differ from the usual goals of photography?
A camera (or post software) must make a single white balance prediction across the frame, not a pixel by pixel choice.
White point estimation is, using a single shot RGB capture, an under-constrained problem.
To calculate the illumination colour we need the surface spectral reflectance properties and the RGB device responses.
We only have the device RGB values.
What if the only illumination in the scene is a decidedly "red" LED? Does it still make sense to think about white-point, or is it better to make the best-informed spectral recreation that we are able to do?
Multi Chanel capture will be better at WB that a 3 channel. A 256 device would be like a spectrophotometer! Good but impractical.
I agree that the more channels, the higher spectral resolution/confidence. So with 256 channels you would know _more_ about the spectrum, with 3 channels you would know _some_ about the spectrum.
If the scene is spectrally "smooth" (and this is known/correctly guessed), then the amount of information to be had by increasing the number of channels should be less. Just like an unsharp lens cannot distinguish (very much) between a high-MP camera and a low-MP camera.
As mentioned by Andrew, a small spectro in the camera at time of capture would help but the problem of multi scene illumination would still arise.
Are you thinking about a spatially averaged, spectrally high-resolution recording, or a more focused "color picker"? If the sky is the brightest surface in the scene (and a prominent part of the image real estate), and the sky has spectral properties that fall within colors that humans on site would classify as "pastel", where would you map this illuminant in the output rendering?
-h
-
It seems to me that, to get h's image-as-open-window effect, what you want to do is duplicate in the image the stimuli that your eyes received at the original scene. In colorimetric terms, if the image can duplicate the same XYZ tristimulus values as the original scene, the cones in your eyes will be stimulated in exactly the same way that they were at the original scene. No need to get all spectral about it, this is basis of color matching.
So the task boils down to reproducing the XYZ tristimulus values of the scene. A standard input ICC profile provides the means of converting the camera RGB values to XYZs. A complication is that color management wants to adapt the raw XYZ numbers to simulate how the colors might appear in an environment with a specific white point, usually D50. To fool the color management system to just let the original raw XYZ values of the scene pass thru without adaptation, I think it would only be necessary to set the camera white balance to D50 (either in the camera settings for JPEG, or in the raw converter when converting raws.) You also need to avoid any other changes to white balance. (There is info at the ICC web (http://www.color.org/scene-referred.xalter) site about this kind of "Scene-Referred" workflow.)
At this point you should have your original scene XYZs intact. Now there is a problem: to display them on a monitor, because monitor white points are usually around D65, the scene XYZ would be shifted bluer by the monitor. So now the need is to perform a conversion to the monitor profile using Absolute Intent. This will adjust the XYZs so that the original scene XYZs are reproduced on the monitor despite its D65 white point.
If I'm not mistaken, this should give you an approximation of the original scene XYZs on a monitor. If the monitor can be taken to the original scene, the colors should appear the same as in the scene, provided the monitor is capable of achieving the same brightnesses as in the original scene. When you have the monitor out in the original scene, you will be viewing the monitor in the same viewing conditions as the scene. It is only under identical viewing conditions that matching XYZ tristimulus values will have the same color appearance.
For the window effect, you will need a monitor capable of matching the brightness of outdoor scenes - it is unlikely you will have one, but high dynamic range monitors are available for $$$. However it is generally ok to scale the XYZs to a lower relative Brightness (which is called Lightness) and still have a good appearance - after all it is usual in photography to display images at other than the original scene brightnesses.
Now as to Torger's OP, the exact matching XYZ color stimuli of a sunset will appear different when viewed under different viewing conditions. Viewing conditions being the color of the light in the viewing environment, the relative luminance of the room light compared to the display, etc. There are Color Appearance Models (such as CIECAM02) that can take the viewing conditions into account and modify the XYZs accordingly in order to have colors appear the same under different viewing conditions. Included are controls for the degree of adaptation, meaning that the models can predict that under stronger colors of light your eyes will not be fully adapted, resulting in a residual color cast in the reproduction. Color Appearance Models haven't really caught on yet in general photography, but might be something interesting to explore.
Cliff
-
So the task boils down to reproducing the XYZ tristimulus values of the scene
This is the most common misunderstanding of the problem. We believe that XYZ values ultimately define how we see colours, while it's a VERY simplified model, limited only to what we perceive when we observe a D50 2 degree stimuli on a grey background in bright surround. In a real world our perception is much more complicated than a simple colour matching function.
-
This is the most common misunderstanding of the problem. We believe that XYZ values ultimately define how we see colours, while it's a VERY simplified model, limited only to what we perceive when we observe a D50 2 degree stimuli on a grey background in bright surround. In a real world our perception is much more complicated than a simple colour matching function.
And unless I'm mistaken, limited only to what we perceive when we observe a D50 2 degree stimuli on a grey background in bright surround of a single color sample.
The answer to the OP's question (the subject line) is simple: No. Not based on the common and affordable solutions used by most people on this forum. IF the question is, Is pleasing color possible in non-standard light, the answer is yes with some work involved.
-
This is the most common misunderstanding of the problem. We believe that XYZ values ultimately define how we see colours, while it's a VERY simplified model, limited only to what we perceive when we observe a D50 2 degree stimuli on a grey background in bright surround. In a real world our perception is much more complicated than a simple colour matching function.
I guess you didn't read the rest of my post.
I would say it's a common misunderstanding to confuse the physical stimulus (XYZ) with the color perception.
XYZ tristimulus values tell you nothing about how a color will be perceived. But if the viewing conditions match, colors with the same XYZ values will be perceived to match in appearance.
XYZ is based on color matching studies and Grassman's laws, and is not limited to D50, grey background, bright surround as you say. To quote from Fairchild's Color Appearance Models
if the signals from the three cone types are equal for two stimuli, they must match in color when seen in the same conditions, since no additional information in introduced within the visual system to distinguish them
-
The three cone types are capable of independent sensitivity regulation, see p. 8
http://www.cis.rit.edu/fairchild/PDFs/AppearanceLec.pdf
So the signals from the three cone types are not equal for two stimuli under different illuminant
-
The three cone types are capable of independent sensitivity regulation, see p. 8
http://www.cis.rit.edu/fairchild/PDFs/AppearanceLec.pdf
So the signals from the three cone types are not equal for two stimuli under different illuminant
Yes, if you change the illuminant, you are either changing the stimuli, or changing the viewing conditions, depending on whether you are talking about reflectances x the illuminant, or about a self-luminous display. The colors generally won't match under either condition, hence the need for CATs and CAMs.
I don't get your point.
-
@ Andrew, post #88
Thanks for your reply. The picture of your POV is clear, but I think it was so even before. It's either colorimetry or nothing at all. No you don't reject the *use* of nothing-at-all, on the contrary, you endorse it; but you do not acknowledge that there could be grades of accuracy within nothing-at-all. Would(n't) this, too, be a correct description of your POV?
I have not used neither 'accurate' nor 'measuring' in this thread. - I remember the very first lesson in physics at high school. The message was: To 'measure' is to 'compare'. There was no requirement of a particular (high) level of precision.
In this sense, when I compare the view of the scene to a stepped scale of white balances, I 'measure' the WB. It is not very precise, but it is subjective only in the sense that the precision may differ from person to person.
Let me expand on the grades of accuracy as I see them.
grade minus 1. Artistic freedom. Consciously deviating from any strict relationship to the perceived colors of the scene.
grade zero. I pull the sliders until I'm happy; I may not be conscious about what I try to achieve: realistic? just pleasing? or I don't care.
grade 1. The attempt to be 'realistic' (to avoid the term 'accurate'). There is one particular view I consciously try to 'hit'. And discover that there are no readily available tools to do it; I'm referred to my memory. This is the grade that Torger has demonstrated here to great effect. The result can not be measured with a colorimeter, but it could easily be validated: everybody who has ever seen this kind of light/scene will at once be able to identify it as being more realistic than e.g. the one obtained by the AWB.
grade 2. This is where I - forgive the vanity - see my own attempt. Like grade 1, with the little addition that the memory factor is eliminated.--edit: And the addition of a stepped scale to aid the visual judgement.
grade 3. Colorimetry. You know all about this one.
But you tell me that -1 to 2 are all the same, because they can not be 'measured' with a spectrometer. I still don't buy it.
-
It's either colorimetry or nothing at all.
That isn't what I wrote. I thought I was very clear that pleasing color which isn't nothing at all is just fine and something most of us strive for and nearly all produce.
I reject the term accuracy in a process that is subjective. How can something subjective have any accuracy metric?
We both go to a location together and setup our cameras side by side. We capture the scene precisely the same moment. We go back to our offices and process the image as we feel we saw the scene. Do they perfectly match? What if they don't? Who's 'correct'? Being subjective, without any means to measure what was there, let alone what happened in our brains, who's to say his capture is more accurate than the other? Using colorimetry, we can easily measure the differences in the two renderings. That doesn't tell us who's closer to the reality of the scene but it tells us how different we processed the data.
I have not used neither 'accurate' nor 'measuring' in this thread.
Good!
- I remember the very first lesson in physics at high school. The message was: To 'measure' is to 'compare'. There was no requirement of a particular (high) level of precision.
That's fine as a start. And it is useful to define the level of precision. But without any measurement, we've put the cart before the horse. How do you measure subjectivity?
In this sense, when I compare the view of the scene to a stepped scale of white balances, I 'measure' the WB. It is not very precise, but it is subjective only in the sense that the precision may differ from person to person.
Agreed. So if we both do this but end up with different results because a large part of the process is subjective, who's more accurate? We can't say. We can argue but we can't come to a firm conclusion. We can ask a 3rd party (who might be biased). They can subjectively tell us which they prefer. None of this helps define which is more accurate in terms of what was there at the scene let alone what we think we remember of the scene. What if you had 10 hours of sleep the night before and I smoked 5 bong loads just before we took the shot <g>? you think that might affect what we remember we saw hours or days later when we process the image to match 'what we recall'?
Let me expand on the grades of accuracy as I see them.
grade minus 1. Artistic freedom. Consciously deviating from any strict relationship to the perceived colors of the scene.
How is the accuracy defined? It isn't.
But you tell me that -1 to 2 are all the same, because they can not be 'measured' with a spectrometer. I still don't buy it.
They are not exactly the same. But they are processes that have no means to measure accuracy. #3 does (or could). Further, what do we do when we use colorimetry to produce a numeric match and it doesn’t look like what we recalled or looks butt ugly? The reality is this is often the case and I illustrated this with actual examples early in this thread. Great, we have a colorimetric match and it looks awful. But it's accurate. The pleasing rendering is, well more pleasing. You can use the old "rate this image on a scale from 1 to 10" but guess what, that's subjective.
Bottom line is asking a question like Is accurate color possible in non-standard light is kind of the wrong question. Because accurate is in the question and the process is so damn subjective.
-
Hi Andrew!
Thank you for your reply.
It is like you neglect the difference between accuracy and precision, and the fact that they are only loosely tied to each other. What you require is not only accuracy, but accuracy AND precision on the spectrometer level.
And I think you exaggerate the degree of imprecision introduced by subjectivity. As I wrote above, I think that "everybody who has ever seen this kind of light/scene will at once be able to identify it as being more realistic than e.g. the one obtained by the AWB."[What I meant was the example in post #23, and it turns out it was not obtained with AWB, but with a grey card].
>hours or days later when we process the image to match 'what we recall'?
You are re-introducing the time-memory factor which is overcome in grade 2.
>So if we both do this but end up with different results because a large part of the process is subjective, who's more accurate? We can't say.
No we can't. But there are 2 Buts:
1-My point is that we are very probably BOTH more accurate than to other persons, who just use a free hand estimate, and even more more accurate than another two persons, who do not do it on location, but are referred to their memory.
And in light like the one under discussion here, all will be more accurate than a WB obtained with the grey card, as shown in post #23, example one. This grey card method is inaccurate not due to poor precision level, but because it is conceptually wrong (for landscape).
2-When we both measure the length of say a football field with a tape measure, the results will differ. Does it mean the term accurate can not be used for the method? I think no. I think it only shows that EVERY measurement has an error, little or large. Our to measurements will differ by decimeters, maybe even meters - but they are very probably BOTH more accurate than a freehand estimate of the length.
That said, I don't really need the word 'accurate' to be happy. I can do with 'realistic' or 'natural'. Only thing I insist on is that grades 1 and 2 are conceptually different from minus one and zero, and that their results are not just arbitrarily pleasing.
They may in fact be less pleasing than those obtained with grades -1 and 0, I agree on that one. If they are, I have different choices: I can 'fix' them or trash them. But at least I'll know approximately 'where I am'.
Too long again!
-
It is like you neglect the difference between accuracy and precision, and the fact that they are only loosely tied to each other. What you require is not only accuracy, but accuracy AND precision on the spectrometer level.
Sounds like those new car commercials (I prefer Sweet and Sour Chicken to just Sour Chicken). So yes, I desire accuracy and precision. The post is about the accuracy of color in non standard light. I didn't write this, I have commented on the issue of using accuracy in the question.
And I think you exaggerate the degree of imprecision introduced by subjectivity. As I wrote above, I think that "everybody who has ever seen this kind of light/scene will at once be able to identify it as being more realistic than e.g. the one obtained by the AWB."[What I meant was the example in post #23, and it turns out it was not obtained with AWB, but with a grey card].
That may be true. It doesn't change the issue of the language the OP is using and the need to express some metric of accuracy which as yet hasn't and probably cannot be stated.
>hours or days later when we process the image to match 'what we recall'?
You are re-introducing the time-memory factor which is overcome in grade 2.
Over come how and if so, by how much?
1-My point is that we are very probably BOTH more accurate than to other persons, who just use a free hand estimate, and even more more accurate than another two persons, who do not do it on location, but are referred to their memory.
Probably but how to prove and by what degree? You know the old saying about opinions.
2-When we both measure the length of say a football field with a tape measure, the results will differ. Does it mean the term accurate can not be used for the method?
It's more accurate than guessing. It's more accurate than using your right foot. We can pull out all kinds of devices, one more accurate than the next to attempt to dismiss the previous measurement. In terms of color accuracy, we at least have a metric of which anything finer is invisible to us. If we measure a color and we are told one instrument says the differences between it and another is a dE of 0.5, another is a difference of 0.05, it's moot. We can't see the difference in either case. It's interesting to know one reference grade device is far more accurate than the other, but if we can't see the lesser accurate difference, who cares?
If your goal is to come up with the length of a football field because you have to build one, does it matter if one measurement is different from the other by 0.05 inches? Or even half a foot? Whoever is paying for the field probably has some metric of error or accuracy they desire. At some point, it doesn't matter. Is +/- 1mm important? With the subjectivity issues I raise, we have no metric to even agree upon as being a useful goal.
I think no. I think it only shows that EVERY measurement has an error, little or large. Our to measurements will differ by decimeters, maybe even meters - but they are very probably BOTH more accurate than a freehand estimate of the length.
But you did measure the item with differing methods to at least come up with a value and one can produce a value of which we agree is acceptable! We do not have this in this discussion because nothing was measured and everything is subjective. If we use your argument above, we have no use for measuring anything because someone can use a higher grade device to disprove the first measurement. If not disprove, suggest it's not accurate. And this is useful to a degree because until we measure and have a metric of what is acceptable, who says the football field has to be within +/- half a foot or 1mm?
That said, I don't really need the word 'accurate' to be happy.
And I said good for you. But there are others here too, and the OP who asked the question using that specific term: accuracy.
But at least I'll know approximately 'where I am'.
How do you express that to others? There's no metric. There's no way to agree with the premise which might be the goal <g>. Again, the question was, is accurate color possible in non-standard light. I submit the question should be reworded. Or we need some means to come up with an accuracy metric. I don't know how you do that without measuring.
-
Only a short reply this time.
>>>hours or days later when we process the image to match 'what we recall'?
>>You are re-introducing the time-memory factor which is overcome in grade 2.
>Over come how and if so, by how much?
Well if I 'measure' the WB at shooting time, I don't rely on memory. By how much depends on the precision of the 'measurement', which is limited, but I believe better than anything I know of.
-
Well if I 'measure' the WB at shooting time, I don't rely on memory. By how much depends on the precision of the 'measurement', which is limited, but I believe better than anything I know of.
Measuring with what, getting what data (CCT which is a range of colors)?
Ever see the CCT values of the camera and the raw converter not agree by a fairly large numeric value? Pretty common for me. I've also found that very often, the reported CCT value and what I end up with after mucking around with Tint/Temp to get a color appearance I prefer is quite different. Are the original values correct or my aim point for color appearance?
-
Measuring with what, getting what data (CCT which is a range of colors)?
Ever see the CCT values of the camera and the raw converter not agree by a fairly large numeric value? Pretty common for me. I've also found that very often, the reported CCT value and what I end up with after mucking around with Tint/Temp to get a color appearance I prefer is quite different. Are the original values correct or my aim point for color appearance?
CCT is measured along the Planckian locus and a more exact color temp includes a tint adjustment. Even so differing WB algorithms will give different results for a given Kelvin value and tint.
Bill
-
@ Andrew, post 103
>Measuring with what, getting what data
In post #99 you had accepted the term 'measurement' for my visual comparison of the scene to camera-recorded white balances:
>>In this sense, when I compare the view of the scene to a stepped scale of white balances, I 'measure' the WB. It is not very precise, but it is subjective only in the sense that the precision may differ from person to person.
>Agreed.
-
It seems to me that, to get h's image-as-open-window effect, what you want to do is duplicate in the image the stimuli that your eyes received at the original scene. In colorimetric terms, if the image can duplicate the same XYZ tristimulus values as the original scene, the cones in your eyes will be stimulated in exactly the same way that they were at the original scene. No need to get all spectral about it, this is basis of color matching.
So the task boils down to reproducing the XYZ tristimulus values of the scene. A standard input ICC profile provides the means of converting the camera RGB values to XYZs. A complication is that color management wants to adapt the raw XYZ numbers to simulate how the colors might appear in an environment with a specific white point, usually D50. To fool the color management system to just let the original raw XYZ values of the scene pass thru without adaptation, I think it would only be necessary to set the camera white balance to D50 (either in the camera settings for JPEG, or in the raw converter when converting raws.) You also need to avoid any other changes to white balance. (There is info at the ICC web (http://www.color.org/scene-referred.xalter) site about this kind of "Scene-Referred" workflow.)
At this point you should have your original scene XYZs intact. Now there is a problem: to display them on a monitor, because monitor white points are usually around D65, the scene XYZ would be shifted bluer by the monitor. So now the need is to perform a conversion to the monitor profile using Absolute Intent. This will adjust the XYZs so that the original scene XYZs are reproduced on the monitor despite its D65 white point.
So what you are saying is: use calibration/profiling to make camera and display behave like (approximations to) ideal XYZ devices. Make sure that the raw WB rendering matches the white-point of the display (by choosing the native WP as WB target, or by calibrating the display WP to D50?).
If I'm not mistaken, this should give you an approximation of the original scene XYZs on a monitor. If the monitor can be taken to the original scene, the colors should appear the same as in the scene, provided the monitor is capable of achieving the same brightnesses as in the original scene. When you have the monitor out in the original scene, you will be viewing the monitor in the same viewing conditions as the scene. It is only under identical viewing conditions that matching XYZ tristimulus values will have the same color appearance.
For the window effect, you will need a monitor capable of matching the brightness of outdoor scenes - it is unlikely you will have one, but high dynamic range monitors are available for $$$. However it is generally ok to scale the XYZs to a lower relative Brightness (which is called Lightness) and still have a good appearance - after all it is usual in photography to display images at other than the original scene brightnesses.
Your example is better than mine was. Place the monitor in the original landscape. Ensure that it is capable of the brightness/gamut that the scene has. Position it in such a way that it covers the same field of view as the image that it is reproducing. Does it _now_ appear as a "window into the landscape"?
(http://image.shutterstock.com/display_pic_with_logo/310684/310684,1236372593,1/stock-photo-lcd-monitor-landscape-background-26174320.jpg)
If it does, then it seems to me that one can be fairly confident that any color issues in reproducing the scene in different contexts (a print hanging on an exhibition wall, the LCD of your laptop at the mall,...) is about context-perceptually motivated re-targeting the rendering (which may or may not be as much "art" as "engineering", but critical to artistic and/or commercial success nonetheless).
My gut-feeling is that even this (granted artificial) exercise can be difficult. I have wrestled some (saturated) reds and greens that cause my camera, raw editor, wide-gamut display, dye printer and profiling to (collectively) behave in unpredictable (for me at least) ways.
I just framed an image in a large, black passepartout. The difference from viewing the image soft-proofed on screen was striking (in a negative way). I do (of course) agree that image surround affects our perception. Rather than treat everything as a "big magic black box that can never be fully understood by humanity", I want to break it down into manageable pieces.
-h
-
Yes, if you change the illuminant, you are either changing the stimuli, or changing the viewing conditions, depending on whether you are talking about reflectances x the illuminant, or about a self-luminous display. The colors generally won't match under either condition, hence the need for CATs and CAMs.
I don't get your point.
The point is that you capture the colour values as if they were under XYZ viewing conditions, and you don't have any information how they should be corrected by CAM.
-
Does it _now_ appear as a "window into the landscape"?
It will never look like a "window in the landscape", it can only more or less remind the original scene (after long & heavy editing). Take a laptop with a decent display, try that "window" trick outdoor, and you'll soon realise that things are far more complicated than you'd wish they were.
-
It will never look like a "window in the landscape", it can only more or less remind the original scene (after long & heavy editing). Take a laptop with a decent display, try that "window" trick outdoor, and you'll soon realise that things are far more complicated than you'd wish they were.
Agreed! If you display scene referred data on the laptop, it's not going to look anything like the scene itself. If you display output referred data on the laptop, well it's been processed by someone or something so we're back to square one (someone or something like the camera itself altering the data). The potentially huge differences between scene gamut and display gamut, contrast ratio and brightness should not be understated either.
-
Hmm…
……….and I thought I was drunk. What are you smooking, you wackos? Mushroom? You should chew it instead.
I want some too. Give it to me now.
-
So what you are saying is: use calibration/profiling to make camera and display behave like (approximations to) ideal XYZ devices. Make sure that the raw WB rendering matches the white-point of the display (by choosing the native WP as WB target, or by calibrating the display WP to D50?).
Yes, effectively using the camera/ICC profile combination as a colorimeter, along with exposure information to obtain absolute XYZ values from the original scene. You don't want color management to alter the XYZ by performing WB or CATs from one assumed white point to another. If white points are kept the same throughout the processing, then color management will have no reason to change anything. Sure, you could calibrate the display to D50, which would simplify things a little.
Your example is better than mine was. Place the monitor in the original landscape. Ensure that it is capable of the brightness/gamut that the scene has. Position it in such a way that it covers the same field of view as the image that it is reproducing. Does it _now_ appear as a "window into the landscape"?
(http://image.shutterstock.com/display_pic_with_logo/310684/310684,1236372593,1/stock-photo-lcd-monitor-landscape-background-26174320.jpg)
If it does, then it seems to me that one can be fairly confident that any color issues in reproducing the scene in different contexts (a print hanging on an exhibition wall, the LCD of your laptop at the mall,...) is about context-perceptually motivated re-targeting the rendering (which may or may not be as much "art" as "engineering", but critical to artistic and/or commercial success nonetheless).
Yes, any changes in viewing conditions (contexts) will result in a different appearance.
As an example, if you don't have a HDR display, you will not be able to reproduce the exact colors (XYZs) of a bright, outdoor scene. Displaying the outdoor image at a much lower luminance on a normal display will never match the original scene due to perceived differences in brightness, contrast and colorfulness. The traditional photographic appearance tweaks to deal with such a situation are to increase the colorfulness (saturation) and the contrast to help make up the difference.
My gut-feeling is that even this (granted artificial) exercise can be difficult. I have wrestled some (saturated) reds and greens that cause my camera, raw editor, wide-gamut display, dye printer and profiling to (collectively) behave in unpredictable (for me at least) ways.
I just framed an image in a large, black passepartout. The difference from viewing the image soft-proofed on screen was striking (in a negative way). I do (of course) agree that image surround affects our perception. Rather than treat everything as a "big magic black box that can never be fully understood by humanity", I want to break it down into manageable pieces.
-h
Although it is difficult to reproduce bright outdoor scenes because of LDR displays, it should be doable in less demanding conditions such as indoor scenes, nightscapes, etc.
Color appearance models break it down into manageable pieces - complicated but manageable.
-
The point is that you capture the colour values as if they were under XYZ viewing conditions, and you don't have any information how they should be corrected by CAM.
A CAM needs the XYZs of whatever it is that you want to match, plus information about the original viewing conditions, and the viewing conditions of the reproduction. There is no limitation of "XYZ viewing conditions," whatever they are. The viewing conditions are whatever they happen to be at the original scene when the camera took the shot, and when viewing the reproduction. The viewing conditions are things like the average luminance, background luminance, XYZs of white, relative brightness of the surround. In order to use a CAM you have to make note of, or estimate the viewing conditions, whatever they are.
-
It will never look like a "window in the landscape", it can only more or less remind the original scene (after long & heavy editing). Take a laptop with a decent display, try that "window" trick outdoor, and you'll soon realise that things are far more complicated than you'd wish they were.
Assuming you laptop doesn't have a HDR display that can match the bright scene luminance, it must have looked dark, something like below while editing on site. Although there was nothing you could do about the brightness difference, you did boost the color and brightness quite a bit, probably just as a CAM would.
Your edited image can't really show anyone what the scene looked like to you, since you edited it to appear the way it did under those specific conditions. Looking at it now on my monitor it can't possibly appear the way it did to you when you edited it.
If you were to repeat your interesting experiment with a scene with luminances that your laptop can actually reproduce, is there a reason you couldn't make it look like a window?
-
Agreed! If you display scene referred data on the laptop, it's not going to look anything like the scene itself. If you display output referred data on the laptop, well it's been processed by someone or something so we're back to square one (someone or something like the camera itself altering the data). The potentially huge differences between scene gamut and display gamut, contrast ratio and brightness should not be understated either.
Those are all good reasons for it not matching. The biggest difference is going to be the brightness. But if you can get a colorimetric match and a brightness match, there is no reason the laptop won't reasonably match the scene under the same viewing conditions.
-
Very interesting discussion, because it makes you question all the things you sort of took for granted.
So here's one thing that I always took for granted: A photograph isn't meant to be an "accurate" representation of reality, in fact that notion is fundamentally flawed to begin with. The whole point is the interpretation, the translation from one realm to another. That is the art of it - what to leave in and what to take out; what to emphasize and what to ignore. The mere act of framing in the viewfinder starts this interpretation and removes from reality.
IOW my conclusion is that "accurate color" is a contradiction in terms. There's no such thing - even if it was technically possible (which I believe it isn't).
---
I'm meeting this head on every day, since I'm a photographer at an art museum and accurate color is the basic requirement that underlies almost everything I do. So I dutifully calibrate and profile and measure everything to the best of my abilities. But I've silently and quietly given up on accuracy a long time ago. What I aim for is equivalent.
-
So here's one thing that I always took for granted: A photograph isn't meant to be an "accurate" representation of reality, in fact that notion is fundamentally flawed to begin with. The whole point is the interpretation, the translation from one realm to another. That is the art of it - what to leave in and what to take out; what to emphasize and what to ignore. The mere act of framing in the viewfinder starts this interpretation and removes from reality.
"Isn't meant to be an "accurate" representation" by whom? The photographer? The camera manufacturer? The raw developer? I'd say that a photograph cannot currently be an "accurate" representation of general physical scenes because (among other things):
-Any movement is frozen
-Focus is fixed
-The rendering is 2d/non-stereoscopic
-DR is limited compared to many scenes
Might it be possible to do accurate reproduction of painted art if there is no visible structure in the paint (and all kinds of expensive, impractical measures are taken, I am sure that you know the trade-offs better than me)? But for general photographers, that might be too limiting.
I don't see why one should not _try_ to make the illusion as "accurate" as possible. It might prove impossible to get all of the way, but getting 90% of the way might be of value.
IOW my conclusion is that "accurate color" is a contradiction in terms. There's no such thing - even if it was technically possible (which I believe it isn't).
If one believes that our perception of vision is only affected by the photons hitting our eyes, then absolute accuracy would consist of merely recording and re-emitting those photons? If we further restrict ourselves to a limited range of wavelengths, a somewhat stochastic/quantized description of those photons (position, energy, etc), the problem is slightly less impractical. So how far do we have to take these simplifications in order to do something that can be made and is still perceptually transparent (or even worse, non-annoying)?
If the OP is often able to get where he wants (impression of print matching memory of scene) by manually tweaking the two common parameters of WB, we seem to be quite close, only needing to find two scalars automatically. Now, if the print was somehow presented as a side-by-side with the original scene, the bar would probably be raised.
-h
-
"Isn't meant to be an "accurate" representation" by whom?
Poor choice of words on my part. What I really mean is "can never be" - and for exactly the reasons you list.
I don't see why one should not _try_ to make the illusion as "accurate" as possible
Me neither, and I do try. But it can never be a perfect match, and that's why I used the term "equivalent". If I hit colors that are equivalent, people will accept that as accurate and be happy with it. And to take it further, that's why it doesn't have to be a perfect match.
-
If one believes that our perception of vision is only affected by the photons hitting our eyes, then absolute accuracy would consist of merely recording and re-emitting those photons?
If they believe that, they need to do a lot more reading on the perception of vision! That's a simplistic view (no pun intended) and far from reality.
-
Some thoughts on some terms.
Accuracy.
@myself post #100
> It is like you neglect the difference between accuracy and precision, and the fact that they are only loosely tied to each other. What you require is not only accuracy, but accuracy AND precision on the spectrometer level.
Well it turns out that my view is the truth of yesterday. It refers to this model: fig 1
(source: http://en.wikipedia.org/wiki/Accuracy_and_precision)
which is now superseded by this one:
fig 2 and 3
In this new terminology, accuracy consists of precision and what is now called trueness, so here precision is in fact part of the definition of accuracy. So noted.
So my goal is to be called 'true' colors from now on.
Subjectivity.
That readings from 2 persons differ, does not make a method subjective. It only means that there is a person-to-person error, which is part of the overall error. This is true for all measurements performed by humans, and something principally different from 'subjective' in the sense in which 'pleasing' is in fact subjective by definition. The statement 'I like this color better' can of course never be measured, validated or falsified.
Trueness could, to some degree. Torger has already given a sketch:
post #54
> say if I have a large number of non-photagraphers with me when I make a shot, and then show them the print I made afterwards and ask them if they think it's an honest reproduction of the scene, if they think this is close to "how it actually looked" and most say yes, then I've succeeded.
This might - theoretically - be improved by choosing painters rather than normal mortals etc.etc.
Good light and 'true' colours!
of course I forgot the attachments:
-
That readings from 2 persons differ, does not make a method subjective.
Not sure I understand what you mean by readings. That sounds like a measurement of some kind so yes, it's not subjective. The perception from two people, fair to call that 'readings'? Fair to say it's not subjective?
-
That was certainly a fast reply!
I was referring to the situation where you and I were applying my 'method' in the same scene. I call the comparison to a stepped scale of white balances for 'reading'. You have earlier agreed to call it 'measurement' of some kind...