Measuring white but has a cast nonetheless
rutt
Registered Users Posts: 6,511 Major grins
Just because you can find a point in your image which you think should be white and it does in fact measure neutral, doesn't mean that your image doesn't have a cast. Here is an image NeiL posted to the WP in this thread.
You'll notice I have inserted a few color sampler points:
The first three points are in the clouds and all do measure a bit blue. The #3 is distinctly blue. #4 is in the sand and it measure slightly purple. The LAB values of these points are easier for me to understand:
Moreover, the evidence of a cast is not just in the neutrals here. Does purple sand make sense? Don't we know that sand is somewhat yellow and not blue? OK, OK, maybe this sand is purple. But purple sand and purple sky are a pattern and should make us guess there is a purple cast. Oh, and our eyes agree. Case closed.
I adjusted by pulling the blue curve to darkness in a luminosity layer:
And now the image looks a lot more believable to me:
In the adjusted image the brightest points in the clouds really do measure neutral, A=0, B=0. And all the blue is out of the sand so it measures a little warm in the shadow and more warm in the sunlight.
The moral: don't look for a single right point. Look at all the evidence. And if your eyes are telling you it's wrong, don't settle for just a little evidence that it is right.
If this were my image, I'd take it through some more steps to enhance contrast and color, but without fixing the cast first, doing this would lead to disaster.
You'll notice I have inserted a few color sampler points:
The first three points are in the clouds and all do measure a bit blue. The #3 is distinctly blue. #4 is in the sand and it measure slightly purple. The LAB values of these points are easier for me to understand:
- L=96, A=3, B=-2
- L=96, A=3, B=-2
- L=91, A-1, B=-4
- L=83, A=6, B=-4
Moreover, the evidence of a cast is not just in the neutrals here. Does purple sand make sense? Don't we know that sand is somewhat yellow and not blue? OK, OK, maybe this sand is purple. But purple sand and purple sky are a pattern and should make us guess there is a purple cast. Oh, and our eyes agree. Case closed.
I adjusted by pulling the blue curve to darkness in a luminosity layer:
And now the image looks a lot more believable to me:
In the adjusted image the brightest points in the clouds really do measure neutral, A=0, B=0. And all the blue is out of the sand so it measures a little warm in the shadow and more warm in the sunlight.
The moral: don't look for a single right point. Look at all the evidence. And if your eyes are telling you it's wrong, don't settle for just a little evidence that it is right.
If this were my image, I'd take it through some more steps to enhance contrast and color, but without fixing the cast first, doing this would lead to disaster.
If not now, when?
0
Comments
There is a reason perhaps for the blue-purple which you find, a reason which exists in the reality of the scene. The tiny amount of blue in the clouds is perhaps reflection from both the sky and the sea, and so should be there.
Sand can be like snow, and as you know the shadows on snow are notoriously blue cast. There is an optical reason for this, too. It is actually seen by the eye. So, maybe should also be there.
In any case, I don't find the very small contribution of these few points excess of blue and red such a problem, mainly because this is what I remember the scene looking like. Your version immediately repels as not what I saw, and to me it looks markedly green.
The atmospheric conditions and light at a scene (and remember this is southern hemisphere Australia), the reflection and refraction that is going on in the complexity of the real world, cannot be totally reduced to post processing orthodoxies. As far as I was concerned there were natural reasons for the magenta tint that you found, and its net effect was rather insignificant, if not enhancing.
Having said that, I don't demand that you should think the same.
But I put it to you, would you want to remove the blush cast - get "proper" LAB numbers - by the explosion of a setting sun (and the sun was in fact setting in this scene, so naturally contributing red to the clouds) just because the whites had more points of red?
http://www.behance.net/brosepix
People and cameras see colors differently. Cameras don't correct casts, people do (at least most people do.) Look at people under tungsten lights. Take a picture under the same conditions (no WP correction). Do the colors look the same to you? Not to most people. That's the reason there is so much fussing about color cast correction. Our visual systems are highly evolved to factor at least some of the light out of the world we live in. When we look at computer monitors (even calibrated ones), out visual systems don't seem to apply the same amount of correction as when we are immersed in reality.
So, I guess I'm hearing your argument as an argument against color cast adjustment, and I think that just won't stand up against the weight of experience. Maybe I'm missing something.
Oh, and end of the day light is usually warmer. It should might be magenta, but it would also be more yellow (though not as much.) Purple sand really just doesn't work.
A prefect example is this:
http://web.mit.edu/persci/people/adelson/checkershadow_illusion.html
What we see (which is the ultimate experience) and what the numbers provide can be vastly different. Instrumentation and the use of numbers is critical in some aspects of imaging and can totally be "wrong" when they are mixed within millions of other values that result in an image.
This isn't all that different in analog photography either. Shoot a scene with a Macbeth on Velvia and Ektachrome, make the grays neutral. Does everything else look the same? Nope. Do you prefer one over the other? Probably. Then that's the correct rendering (even though colorimetrically, neither is correct).
Author "Color Management for Photographers"
http://www.digitaldog.net/
I just remeasured the wave foam and I get 1 to 3 points more blue (1%C 1%M 0%Y), so yes there is a blue cast, and it has been introduced by my post processing. The question remains, is that very small imbalance the main reason why the scene looks purplish? I don't think so. I think this scene is naturally purplish.
Our eyes adjust for color cast whites, but they are not blind to color casts, otherwise we would have no admiration for sunsets!
http://www.behance.net/brosepix
So, while I usually say don't trust your eyes, in this particular case, I have a different point: don't trust a single measurement. Also, don't ask for critique and then argue with it.
Ultimately its your image so you have to render as you prefer. The top image does appear magenta but I prefer it to the overly green (to my eye) rendering below it. I'm sure lots of people could find a common ground here. Point is, neutralize any number of areas and that doesn't make the image correct. I find when I WB in Lightroom, I almost always then slide the tint or temp slider to warm it up. Again, colorimetrically its may not be right, even when there's a true neutral gray (Macbeth or similar target) in the scene. What's right isn't what one set of numbers tells you but what your eyes and aesthetics tell you is correct.
Author "Color Management for Photographers"
http://www.digitaldog.net/
rutt, I edited my post above yours, please look back.
I don't argue, but boy do I discuss. I listen with gratitude, I consider. When it comes to any image the photographer's point of view is as valid as anyone else's. It is not necessary for the person asking for critique to go into suspended animation for critique to work. Critique to me is respectful dialog with the aim of everyone progressing in some way or other. A good example is this very thread.
I have complimented you on the point you raise. I have gone back to my image. I have confirmed the color cast. I have assessed it and your variation. My inclination now is to get that wave foam neutral. It will make a very tiny difference to the whole image, which I think nature has made purplish, but I will feel better.
You ask more of me than that??!!
http://www.behance.net/brosepix
http://www.behance.net/brosepix
As to fixing the cast, a big part of my point is that you cannot just find something you think should be white and click on it and expect to get rid of all the cast every time in all your pictures. There can be mixed casts, light reflected and with a different color from the direct light. There can be sources of light which don't share the overall cast of the image. There are all sorts of reasons for casts in the highlights but not in the shadows or visa versa. LR's single click WB is a great easy tool, but it's not as powerful as separate channel curves which in turn aren't as powerful as layers and layer masks...
I'm suspicious of balancing on the water if it leaves the purple in the sand and whitest parts of the clouds. I'm not sure I understand the relationship of the white foam to the overall lighting of the scene. I don't think it's like a gray card, reflecting that lighting without necessarily coloring it.
You are repeating some of my own comments.:D
A "cast" is not always something foreign to an image and to be got rid of.
I repeat, this scene was made purplish mainly by nature. I see it all the time back home.
Yes, white is made by mixing all colors, and mix all colors is what sea foam does. It should be neutral. Even the foam of muddy water, if the droplets are small enough, is pure white. I think!:D
http://www.behance.net/brosepix
I am left with a question.
John, if the sand was more magenta than it was blue ( L=83, A=6, B=-4 ) , why did you only correct the blue cast ( b=-4), and ignore the magenta(A=6)?
I like the warmer second version, and I am sure there is more yellow in the greenery as a result of the decrease in the blue. Vegetation is typically higher in yellow than green or B is positive and > than A is negative.
I am certain that clouds are not always white. Clouds exhibit traces of the color of the landscape/seascape reflected from beneath them. Slightly red over red deserts in the SouthWest, and slightly blue over deep ocean water....
I am also certain that my monitor has been recently calibrated with a new i1D2, courtesy of Andrew's recommendation. I am certain his monitor has been also.
Moderator of the Technique Forum and Finishing School on Dgrin
I thought that sand should be warm. Not sure that it shouldn't be magenta. Just sure that it shouldn't be blue. In fact, I was just considering pulling the green curve a little toward darkness to add a little more magenta in response to NeiL and Andrew who both saw my image as too green (not magenta enough.) NeiL could always do that if he cared.
That's true enough. But the whitest parts of these clouds could be white or very close and the darker parts could still be a little blue where the sky comes through. That's actually how I left it.
I never really thought otherwise.
Even supposing that's true, does it imply that the water reflects enough of the ambient light neutrally to be an appropriate point to balance on.
Not necessarily in either case. Point #1:http://www.ppmag.com/reviews/200512_rodneycm.pdf
Point #2: The foam would not be white based on a number of factors (the time of day, the reflectivity of the sky and other nearby objects). Snow is "supposed to be white" and yet, its nearly always bluish in say noontime daylight for a number of reasons. You can make it totally neutral numerically for an effect, but other areas in the scene can be negatively affected (negatively in terms of a preferred rendering which is subjective).
Lastly, it maybe should not be neutral depending on the intent of the image creator. This is after all, output referred imagery (the distinction is important and defined here:http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf
Author "Color Management for Photographers"
http://www.digitaldog.net/
Thanks very much Andrew, I will follow this up as I get the chance!
http://www.behance.net/brosepix
Very interesting to read your comments, thanks!
One reaction I have, which is similar to things Andrew is pointing out, is that a magenta cast might have as much justification to remain unmolested in a photograph as the blush of a sunset, even to be intensified like sunsets are. It was mainly the natural cast of light in this photo, the artificial color cast my processing introduced is having a secondary effect, I believe. I will see when I neutralise the sea foam.
Sand can have many hues, complicated by the color of the light on it. In the location of the photo the sand looks white to the eye in real life, much closer to snow than earth, and more so in reduced light such as here. It is a cool light time of a winter day, late afternoon but before a sunset proper. The strongest light on the sand is the intense blue from the sky. I wonder why you would want to make the sand yellow-brown-green???
Not all vegetation is predominantly yellow. In this photo where it is predominantly yellow it is predominantly yellow. Otherwise it is not. Why would you want to give it a color it doesn't have.
If I were to be a little naughty I would suggest that some people use Margulis as a security pillow to avoid the challenges of real light
But seriously I do agree that there is a light challenge in my shot which I have not successfully met, yet. With your fantastic input I have a better chance!!
http://www.behance.net/brosepix
Tai Jimenez, Boston Ballet's Production of Balanchine's Serenade
Oh, how I hate to have to post one of my own favorite shots (again) in order to make a point.
Anyway, here is an example of an image which needs to show blue ambient light. Serenade has a traditional staging with these white costumes and this blue light. We know the costumes are white, but use the WB dropper on them and disaster ensues. At the very best, we lose the blue light and fail to convey the mood of the dance. In fact, we get very unrealistic flesh tones as well; way too hot. The light is so blue that it overcomes our visual systems' simultaneous contrast and we do see the scene with a blue cast. The lighting designer has made sure of that.
Leaving the scene the way it was or using some sort of auto white balance is also not an option. I no longer have the raw version of this, but suffice to say, every simple option results in an image which is very blue, unbelievably so. It just doesn't look right at all.
So here is what I did. I decided that some of her flesh should measure within the conventional rule of thumb, i.e., A and B both positive and B>=A. Not all, but some. As you can see some of her is very blue. The costume is very blue. But our visual system is powerful enough to be able to overcome the blue light just enough to tell us that she is healthy and has good skin tone. So I found some places where the original measured closest to the rule and adjusted the image to make those follow the rule. Once I did this, the image looked great to my eye, showing the blue cast while maintaining believability.
My point? Actually I have two:
Here is a suggestion for NeiL's image. Make a layer in PS and use curves with color blend to neutralize (say) the whitest point of the clouds (like the point where I had color sampler #1). Then play with the opacity of that layer to taste. Or make the sand warm and again play before playing the opacity. I predict that if you work with the visual system instead of fighting it, you will have more success in showing the color of the light instead of an image with an obvious cast crying out to be fixed.
Overall, it has a green "look" that is less pleasing to me visually. Usually, if one has to side between "too green" or "too magenta", most people I believe would pick the later. I think the original IS too magenta but the fix is too green. I'd split the difference.
One thing that is useful is to "work hot" and over do the correction, the use the Fade command (assuming you're doing this in Photoshop) to fine tune an effect. Of course you can do this with adjustment layers too. Fade is super useful for painting moves. Point is, work hotter (more extreme) than you think the adjustment should be, then fade back. In the cases above, both need a fade back IMHO.
Author "Color Management for Photographers"
http://www.digitaldog.net/
@ rutt Yes I remember the series your shot above of the dancer came from. Actually I bookmarked your whole post in which you describe your processing of them. They make your case very strong. However, it's not possible to examine the pixel quality of them, and as you know that would be an issue for me!
http://www.behance.net/brosepix
Very relevant. I even said that my internal AWB is more diligent than most other people's.
I didn't know about this one, and thanks for it!
The post I was referring to is:
http://www.dgrin.com/showpost.php?p=206778&postcount=9
http://www.behance.net/brosepix
A "black body" absorbs radiation and re-emits it as energy with its own signature depending on the nature of that body. I doubt that different black bodies could produce the exact same net energy output, in other words different blackbodies would have different RGB values.
Bodies which generate energy, maybe your "luminants", also emit their energy in a characteristic way depending on their composition, and described by spectroscopy. In this case, probably the net radiation from different sources could be equal (though their spectroscopy profiles would still be different, and therefore again their RGB values would be different).
Objects which are are not energetic enough to produce blackbody radiation, and which do not generate enough energy to be "luminants", are absorbers and/or reflectors.
I would put water in the breaking wave state in this last category, and I would guess that it is overwhelmingly a reflector. I would also guess that its range of reflective particle sizes, from microscopic upwards, would result in all the frequencies of incident light being reflected and mixed - into white! Pure white! Equal RGB values!
http://www.behance.net/brosepix
D65 is a spectral power distribution (a certain amount of energy at each wavelength across the visible
spectrum).
b) D65 is a tristimulus value; the D65 spectrum, when viewed by the CIE standard observer, produces an
XYZ triplet (or xyY if you prefer).
c) 6500 K blackbody radiator is a spectral power distribution.
d) 6500 K is blackbody tristimulus value; the 6500 K blackbody spectrum, when viewed by the CIE standard
observer, produces an XYZ triplet—similar to, but slightly different from, the one found in (b).
e) Correlated color temperature takes a color’s chromaticity coordinate (x,y) and finds the particular blackbody
temperature whose chromaticity coordinate (d) is closest to it. Note that there are many different
colors that have the same correlated color temperature. So a spectrum is very precise and unique. Its
xyY is less precise and unique. Its CCT is even less precise and unique.
D65 is a unique SPD (there exists only one). A color whose CCT is 6500 K is not unique (there are infinitely
many different xyY and SPDs that share it).
This is D65, as it falls on the Black Body curve:
Author "Color Management for Photographers"
http://www.digitaldog.net/
http://www.behance.net/brosepix
I think this is agreeing with me, no?
What we need now is a nice closing statement which links all of this relevantly to my image, and in particular to the "whiteness" of the breaking wave.
I have done that. That was the purpose of my post #26 above.
http://www.behance.net/brosepix
I don't know that any of this proves that the wave or spray should be white (or that white is equal amounts of RGB). If we agree its a reflector, it doesn't necessarily reflect a neutral or white. At least at sunset and throughout many times of the day (and based on the conditions of the weather on that day). Or objects nearby the spray that reflect other colors.
Author "Color Management for Photographers"
http://www.digitaldog.net/
http://www.behance.net/brosepix