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.
Point taken, but in the case of my image I think there are enough of the wavelengths in the spectrum present in the incident light on the water to get pretty close to white. Indeed, your technical text suggests we are always dealing with approximations.
You do also make the point that it doesn't really matter whether we white balance on "actual" white, so long as we like the result. There is truth in that, certainly. But then what is there to discuss in that case?
Another interesting phenomenon is that while we are seeing a scene, such as in my image, live, we may see a blue cast because the sky is blue. However, in fact the sky is more violet than blue, but our eyes are less sensitive to the violet wavelength.
"Violet" and "Blue" are words we use to describe color and color is something that we perceive in our brains. That is, color, is a perceptual property. If we can't see it it's not a color. Therefore, color is not a particular wavelength of light. It is a cognitive perception that is the end result of the excitation of photoreceptors followed by retinal processing and ending in the visual cortex.
We define colors based on perceptual experiments.
A coordinate in a "colorspace" outside the spectrum locus is not a color. We often refer to these as "imaginary colors" but this is by and large also erroneous (you can't map an imaginary color from one
colorspace to another as the math (and experimental data) for each color space breaks down outside the spectrum locus.
"Violet" and "Blue" are words we use to describe color and color is something that we perceive in our brains. That is, color, is a perceptual property. If we can't see it it's not a color. Therefore, color is not a particular wavelength of light. It is a cognitive perception that is the end result of the excitation of photoreceptors followed by retinal processing and ending in the visual cortex.
We define colors based on perceptual experiments.
A coordinate in a "colorspace" outside the spectrum locus is not a color. We often refer to these as "imaginary colors" but this is by and large also erroneous (you can't map an imaginary color from one
colorspace to another as the math (and experimental data) for each color space breaks down outside the spectrum locus.
Does it follow then that we don't have a meaningful word in "color"? Something which is 100% subjective and nonreferential cannot be communicated.
A coordinate in a "colorspace" outside the spectrum locus is not a color. We often refer to these as "imaginary colors" but this is by and large also erroneous (you can't map an imaginary color from one
colorspace to another as the math (and experimental data) for each color space breaks down outside the spectrum locus.
True, but what is its relevance to a discussion of white balance?
"Violet" and "Blue" are words we use to describe color and color is something that we perceive in our brains. That is, color, is a perceptual property. If we can't see it it's not a color. Therefore, color is not a particular wavelength of light. It is a cognitive perception that is the end result of the excitation of photoreceptors followed by retinal processing and ending in the visual cortex.
We define colors based on perceptual experiments.
A coordinate in a "colorspace" outside the spectrum locus is not a color. We often refer to these as "imaginary colors" but this is by and large also erroneous (you can't map an imaginary color from one
colorspace to another as the math (and experimental data) for each color space breaks down outside the spectrum locus.
Doesn't the phenomenon I described in post #28 explain why the footprint in the snow (photo in a link) is violet?
You don't really mean that, do you? Music does it all the time.
When I started it, I thought this would be sort of an open and shut case. The image has a blue cast. I adjusted to get the blue out of the clouds and sand. OK, so NeiL the photographer hated the adjustment. Fine. It's his shot. He could find some in-between sort of adjustment or not. I proposed a way to think about showing colored light so that it doesn't look like a cast. OK, again, NeiL can take it or leave it.
Then there is an issue about on-click white balance vs curves. It was my point, so I understood it, even if nobody else did.
But then it got way beyond me and has now reached heights so refined that I'm way too stupid to follow.
I'm just hoping to see a version of the picture that NeiL likes which doesn't have the horrible cast.
Point taken, but in the case of my image I think there are enough of the wavelengths in the spectrum present in the incident light on the water to get pretty close to white. Indeed, your technical text suggests we are always dealing with approximations.
"Digital photographic" white is not equal RGB???
They are totally approximations unless you measure the values, in something like CIE/XYZ or Lab using something like a Spectrophotometer. That's why I posted the link about scene versus output referred. And no, equal values of RGB don not always make a neutral. They do in well behaved theoretically built RGB working spaces. There are all kinds of device dependent RGB spaces where R=G=B isn't neutral. The RGB values could be 255/245/248 and it would be quite easy to alter the last two values to 255. That doesn't make the scene referred values a neutral white highlight. RGB isn't device independent (Lab is). If you have a measurable Lstar value of 100, 0a/0b, that's a neutral white.
"Pretty close to white" is in itself, an approximation.
When I started it, I thought this would be sort of an open and shut case. The image has a blue cast. I adjusted to get the blue out of the clouds and sand. OK, so NeiL the photographer hated the adjustment. Fine. It's his shot. He could find some in-between sort of adjustment or not. I proposed a way to think about showing colored light so that it doesn't look like a cast. OK, again, NeiL can take it or leave it.
Then there is an issue about on-click white balance vs curves. It was my point, so I understood it, even if nobody else did.
But then it got way beyond me and has now reached heights so refined that I'm way too stupid to follow.
I'm just hoping to see a version of the picture that NeiL likes which doesn't have the horrible cast.
"Don't worry, be happy!" Now who said that earlier...
But then, if your aim is to strip people of their opinions then I guess you will just have to be unhappy...wink
And, please, be careful whose cast you're calling horrible, G...d!cry:D
I don't believe you are in splendid isolation in understanding what you are doing, it's simply that I haven't found your results in this case preferable, and I am suspicious of side effects. I have seen some truly horrible stuff from LABrolleyes
On the other hand, as a result of this dialog with you I am somewhat more inclined to reconsider LAB. You might have to be satisfied with that for the moment.
I'll send you a PM later today. Whole separate topic than what color sand and surf ought to be. I didn't mean to derail this interesting thread, and I apologize for that.
They are totally approximations unless you measure the values, in something like CIE/XYZ or Lab using something like a Spectrophotometer. That's why I posted the link about scene versus output referred. And no, equal values of RGB don not always make a neutral. They do in well behaved theoretically built RGB working spaces. There are all kinds of device dependent RGB spaces where R=G=B isn't neutral. The RGB values could be 255/245/248 and it would be quite easy to alter the last two values to 255. That doesn't make the scene referred values a neutral white highlight. RGB isn't device independent (Lab is). If you have a measurable Lstar value of 100, 0a/0b, that's a neutral white.
Are we still on the same topic?
I have been suggesting that the most neutral item in my pic is likely to be the breaking wave, and so if click white balance/neutralise white is going to be the way to deal with the color cast that is the best place to do it.
rutt has been arguing for selective WB via LAB (but still using a "neutral' reference), and I recognise that on the face of it his method would seem to have more sophistication. However, I am not convinced yet that it can produce the best results in terms of pixel quality ie minimal degradation of pixel data and aesthetic effect.
I'll send you a PM later today. Whole separate topic than what color sand and surf ought to be. I didn't mean to derail this interesting thread, and I apologize for that.
Oh, R, it's Sunday morning here (ie Friday morning) and I'm "feelin' groovy", I don't give a fig what you derail. Carry on, do your thing! I am a million miles from getting uptight about it! Andrew and rutt are on quite a trip. Bubbles are a big temptation!
rutt has been arguing for selective WB via LAB (but still using a "neutral' reference), and I recognise that on the face of it his method would seem to have more sophistication. However, I am not convinced yet that it can produce the best results in terms of pixel quality ie minimal degradation of pixel data and aesthetic effect.
Not at all. I never took the image to LAB for the adjustment I posted. I did it with RGB curves on a color blend layer in RGB. The dance picture did make a trip to LAB, but forget that; I only posted it as an example of showing colored light in a believable way.
I have a simple point, once I think should be uncontroversial. One-click-white-balance is less powerful than RGB curves as a way to adjust color balance. The reason is pretty simple. With curves, one can apply different (less, no) adjustment to darker (more or less red, green, blue) areas. One can adjust an image without any known neutral point, for example, by using flesh, sky, greenery, sand, or a combination as clues.
By using color blend mode for the curve layer, one restricts the impact to color and it has no impact on contrast. I suppose that is LAB-like, but it differs from LAB curves. To get the same effect with LAB curves, one would need multiple curve layers with layer masks or blend-if restrictions.
Nobody has to agree with me. I love to be proven wrong and to learn new techniques. But I am a kind of concrete guy and like to see theory applied. In this way, I guess I'm sort of in agreement with Andrew's theory. We can reason all we like. We can use any technique we can dream up. In the end all that really matters is our reaction to the end product. I'd like to see a version the photographer approves of.
I have been suggesting that the most neutral item in my pic is likely to be the breaking wave, and so if click white balance/neutralise white is going to be the way to deal with the color cast that is the best place to do it.
In this picture, I'd agree its a good starting point to click balance and then modify to taste. In any other image, YMMV (depends on all kinds of factors).
Should the shadows be cooler than the sunlit areas? Visa versa? How much so?
Should the tops of the clouds be cooler than the bottoms? Visa versa? How much so?
How much green in the blue sky?
Do you want some sunset colors in the sky clouds (warmer)?
Lots of different adjustments will keep that white water neutral without addressing the above. I don't believe you are going to nail this image without paying attention to these questions somehow or another.
Should the shadows be cooler than the sunlit areas? Visa versa? How much so?
Should the tops of the clouds be cooler than the bottoms? Visa versa? How much so?
How much green in the blue sky?
Do you want some sunset colors in the sky clouds (warmer)?
Lots of different adjustments will keep that white water neutral without addressing the above. I don't believe you are going to nail this image without paying attention to these questions somehow or another.
Well, I wouldn't want to diminish the authority of the actual light in the scene too much.
As I said before, I am not really enthusiastic about losing touch with reality in this image. I get the feeling some PP activities are a way of avoiding the challenges of real light by cutting holes in an image and filling them back in with controllable artificiality, perfectly made according to our own prejudices. Now, artificiality gets a lot of easy applause...
Also, as you know, I am uneasy using the slider degradations. I think they lack sympathy for the data in the pixels.
In this picture, I'd agree its a good starting point to click balance and then modify to taste. In any other image, YMMV (depends on all kinds of factors).
Yes.
However it has been pointed out to me that specular highlights should probably not be used for this purpose, and that the breaking wave might be too close to a specular highlight.
Yes.
However it has been pointed out to me that specular highlights should probably not be used for this purpose, and that the breaking wave might be too close to a specular highlight.
If done with a raw converter, you'd want something a bit lower in tone than a specular (and ACR/LR will pop a warning if its too hot). Something like the 2nd white patch on the Macbeth. In a gamma corrected image, you'd want to gray balance. In Raw, you'd want to work closer to a highlight due to how the data is distributed (linear encoded).
If done with a raw converter, you'd want something a bit lower in tone than a specular (and ACR/LR will pop a warning if its too hot). Something like the 2nd white patch on the Macbeth. In a gamma corrected image, you'd want to gray balance. In Raw, you'd want to work closer to a highlight due to how the data is distributed (linear encoded).
Here is a version which acknowledges, I think, the out-casters' complaints, without moving too far away from my experience of this sight. Among other things, I used a curves adjustment on the blue channel (in the opposite direction to what your screenshot on page 1 shows , rutt???) which resulted in some combing in the histogram which thrilled me, as you can imagine! I saw no combing in your version of the image on page 1, rutt??? Why mine and not yours???
I think now, that I would really prefer to do these color cast adjustments in LAB rather than in RGB, if they are able to preserve better pixel quality.
So, this has turned out to be a rather significant thread in the development of my post processing attitude, approach and skills. Thanks to you both, rutt and Andrew.
What I need now is a little hand to get started on this kind of thing in LAB. I guess I can first study the links I have in this thread. What do you think? Is there a LAB primer? I would rather go via Easy Street than Margulis's Wormhole!!!
Here is a version which acknowledges, I think, the out-casters' complaints, without moving too far away from my experience of this sight.
Better. Now its really a judgement call based on the image creator and how they wish to express the scene they recall (output referred).
What I need now is a little hand to get started on this kind of thing in LAB.
Why? You "fixed" the issue. How's tossing away a lot of bits and spending time waiting on conversions in any way beneficial based on what you did? Ideally all this would be done in the Raw converter anyway. Faster, no data loss, etc.
Looks much better, NeiL. Don't you agree. For fixing this kind of cast, RGB curves in a color blend layer is probably what you want. LAB is great, and can fix much worse casts quickly, but these days I always start out doing just this to fix minor casts.
BTW, adjustments to the blue curve don't tend to impact contrast, but the other channels do. It's a good habit to use a color blend layer to ensure that all you change is color not contrast.
Comments
Point taken, but in the case of my image I think there are enough of the wavelengths in the spectrum present in the incident light on the water to get pretty close to white. Indeed, your technical text suggests we are always dealing with approximations.
"Digital photographic" white is not equal RGB???
http://www.behance.net/brosepix
http://www.behance.net/brosepix
"Violet" and "Blue" are words we use to describe color and color is something that we perceive in our brains. That is, color, is a perceptual property. If we can't see it it's not a color. Therefore, color is not a particular wavelength of light. It is a cognitive perception that is the end result of the excitation of photoreceptors followed by retinal processing and ending in the visual cortex.
We define colors based on perceptual experiments.
A coordinate in a "colorspace" outside the spectrum locus is not a color. We often refer to these as "imaginary colors" but this is by and large also erroneous (you can't map an imaginary color from one
colorspace to another as the math (and experimental data) for each color space breaks down outside the spectrum locus.
Author "Color Management for Photographers"
http://www.digitaldog.net/
http://www.behance.net/brosepix
Does it follow then that we don't have a meaningful word in "color"? Something which is 100% subjective and nonreferential cannot be communicated.
http://www.behance.net/brosepix
http://www.behance.net/brosepix
True, but what is its relevance to a discussion of white balance?
http://www.behance.net/brosepix
Doesn't the phenomenon I described in post #28 explain why the footprint in the snow (photo in a link) is violet?
http://www.behance.net/brosepix
You don't really mean that, do you? Music does it all the time.
When I started it, I thought this would be sort of an open and shut case. The image has a blue cast. I adjusted to get the blue out of the clouds and sand. OK, so NeiL the photographer hated the adjustment. Fine. It's his shot. He could find some in-between sort of adjustment or not. I proposed a way to think about showing colored light so that it doesn't look like a cast. OK, again, NeiL can take it or leave it.
Then there is an issue about on-click white balance vs curves. It was my point, so I understood it, even if nobody else did.
But then it got way beyond me and has now reached heights so refined that I'm way too stupid to follow.
I'm just hoping to see a version of the picture that NeiL likes which doesn't have the horrible cast.
Actually, RGB values alone don't tell us what a color looks like.
Author "Color Management for Photographers"
http://www.digitaldog.net/
They are totally approximations unless you measure the values, in something like CIE/XYZ or Lab using something like a Spectrophotometer. That's why I posted the link about scene versus output referred. And no, equal values of RGB don not always make a neutral. They do in well behaved theoretically built RGB working spaces. There are all kinds of device dependent RGB spaces where R=G=B isn't neutral. The RGB values could be 255/245/248 and it would be quite easy to alter the last two values to 255. That doesn't make the scene referred values a neutral white highlight. RGB isn't device independent (Lab is). If you have a measurable Lstar value of 100, 0a/0b, that's a neutral white.
"Pretty close to white" is in itself, an approximation.
Author "Color Management for Photographers"
http://www.digitaldog.net/
OK, give us an example
http://www.behance.net/brosepix
"Don't worry, be happy!" Now who said that earlier...
But then, if your aim is to strip people of their opinions then I guess you will just have to be unhappy...
And, please, be careful whose cast you're calling horrible, G...d!
I don't believe you are in splendid isolation in understanding what you are doing, it's simply that I haven't found your results in this case preferable, and I am suspicious of side effects. I have seen some truly horrible stuff from LAB
On the other hand, as a result of this dialog with you I am somewhat more inclined to reconsider LAB. You might have to be satisfied with that for the moment.
http://www.behance.net/brosepix
I'll send you a PM later today. Whole separate topic than what color sand and surf ought to be. I didn't mean to derail this interesting thread, and I apologize for that.
That's what I said.
Are we still on the same topic?
I have been suggesting that the most neutral item in my pic is likely to be the breaking wave, and so if click white balance/neutralise white is going to be the way to deal with the color cast that is the best place to do it.
rutt has been arguing for selective WB via LAB (but still using a "neutral' reference), and I recognise that on the face of it his method would seem to have more sophistication. However, I am not convinced yet that it can produce the best results in terms of pixel quality ie minimal degradation of pixel data and aesthetic effect.
http://www.behance.net/brosepix
Oh, R, it's Sunday morning here (ie Friday morning) and I'm "feelin' groovy", I don't give a fig what you derail. Carry on, do your thing! I am a million miles from getting uptight about it! Andrew and rutt are on quite a trip. Bubbles are a big temptation!
http://www.behance.net/brosepix
I made this same point in "Aussie triplets"!
http://www.behance.net/brosepix
Not at all. I never took the image to LAB for the adjustment I posted. I did it with RGB curves on a color blend layer in RGB. The dance picture did make a trip to LAB, but forget that; I only posted it as an example of showing colored light in a believable way.
I have a simple point, once I think should be uncontroversial. One-click-white-balance is less powerful than RGB curves as a way to adjust color balance. The reason is pretty simple. With curves, one can apply different (less, no) adjustment to darker (more or less red, green, blue) areas. One can adjust an image without any known neutral point, for example, by using flesh, sky, greenery, sand, or a combination as clues.
By using color blend mode for the curve layer, one restricts the impact to color and it has no impact on contrast. I suppose that is LAB-like, but it differs from LAB curves. To get the same effect with LAB curves, one would need multiple curve layers with layer masks or blend-if restrictions.
Nobody has to agree with me. I love to be proven wrong and to learn new techniques. But I am a kind of concrete guy and like to see theory applied. In this way, I guess I'm sort of in agreement with Andrew's theory. We can reason all we like. We can use any technique we can dream up. In the end all that really matters is our reaction to the end product. I'd like to see a version the photographer approves of.
In this picture, I'd agree its a good starting point to click balance and then modify to taste. In any other image, YMMV (depends on all kinds of factors).
Author "Color Management for Photographers"
http://www.digitaldog.net/
- Should the shadows be cooler than the sunlit areas? Visa versa? How much so?
- Should the tops of the clouds be cooler than the bottoms? Visa versa? How much so?
- How much green in the blue sky?
- Do you want some sunset colors in the sky clouds (warmer)?
Lots of different adjustments will keep that white water neutral without addressing the above. I don't believe you are going to nail this image without paying attention to these questions somehow or another.No problem, your intention has always been absolutely clear.
I'll get back to the image.
http://www.behance.net/brosepix
Well, I wouldn't want to diminish the authority of the actual light in the scene too much.
As I said before, I am not really enthusiastic about losing touch with reality in this image. I get the feeling some PP activities are a way of avoiding the challenges of real light by cutting holes in an image and filling them back in with controllable artificiality, perfectly made according to our own prejudices. Now, artificiality gets a lot of easy applause...
Also, as you know, I am uneasy using the slider degradations. I think they lack sympathy for the data in the pixels.
So I am feeling my way cautiously...
http://www.behance.net/brosepix
Yes.
However it has been pointed out to me that specular highlights should probably not be used for this purpose, and that the breaking wave might be too close to a specular highlight.
http://www.behance.net/brosepix
If done with a raw converter, you'd want something a bit lower in tone than a specular (and ACR/LR will pop a warning if its too hot). Something like the 2nd white patch on the Macbeth. In a gamma corrected image, you'd want to gray balance. In Raw, you'd want to work closer to a highlight due to how the data is distributed (linear encoded).
Author "Color Management for Photographers"
http://www.digitaldog.net/
Another piece to the puzzle!
Thanks.
http://www.behance.net/brosepix
I think now, that I would really prefer to do these color cast adjustments in LAB rather than in RGB, if they are able to preserve better pixel quality.
So, this has turned out to be a rather significant thread in the development of my post processing attitude, approach and skills. Thanks to you both, rutt and Andrew.
What I need now is a little hand to get started on this kind of thing in LAB. I guess I can first study the links I have in this thread. What do you think? Is there a LAB primer? I would rather go via Easy Street than Margulis's Wormhole!!!
Neil
http://www.behance.net/brosepix
And when I update the histogram cache for my image above the combing disappears! Magic!!
No mystery, then!
http://www.behance.net/brosepix
Better. Now its really a judgement call based on the image creator and how they wish to express the scene they recall (output referred).
Why? You "fixed" the issue. How's tossing away a lot of bits and spending time waiting on conversions in any way beneficial based on what you did? Ideally all this would be done in the Raw converter anyway. Faster, no data loss, etc.
Author "Color Management for Photographers"
http://www.digitaldog.net/
BTW, adjustments to the blue curve don't tend to impact contrast, but the other channels do. It's a good habit to use a color blend layer to ensure that all you change is color not contrast.