colour channels in the histogram and 'good' white balance

[stuck]

At the risk of starting yet another white balance thread...

I'm still nothing more than a rank beginner in this PP lark but as I've been tinkering, I'm wondering if one way of knowing that the colour balance is close to correct is to look at the alignment of the colour channels on the histogram. 'Right' being when they all more or less line up/overlap, as opposed 'wrong' when one channel is offset relative to the others.

For example in a picture where 'as shot' equates to 5500 then in ACR the R & G channels are more or less coincident so they are plotted as a white 'bump' but the blue channel is offset such that a large proportion of it appears as a blue bump at the right hand end of the histogram. Unsurprisingly the picture as a blue cast.

Choosing 'auto' for the colour temp raises it to 7200 and the blue slips left leaving the right of the histogram shaded in yellow, because that's where the R & G are. Now the picture looks a bit warm.

Picking 'cloudy' lowers the temp to 6500 and now the histogram is more or less all white, i.e. all colour channels have a more or less similar profile. Visually, on a not very good LCD, this option looks 'good'.

Have I discovered a guiding principle that makes for sensible starting point in PP that everyone else knows already because they've been at PP for years and/or read tutorials and books that I've not?

Ken

[DanK]

Nope. the positions of the 3 histograms depend on the mix of colors in the image. Imagine two images. One is all red except for a small circle of neutral white. the second is all blue except for a small circle of neutral white. Proper white balance would render both circles identically, but the first image would then have a huge red histogram and almost nothing else, while the second would be almost all blue.

If you want accurate white balance, shoot raw, put a good spectrally neutral card somewhere in one image (I use a whiBal), and set white balance for the card. Let the histograms fall wherever they may.

[Mike Buckley]

I can confirm Dan's post that alignment of the color channels has nothing to do with accurate white balance.

[Steve S]

Alignment of the right hand side , is a neutral WB.

[stuck]

I can confirm Dan's post that alignment of the color channels has nothing to do with accurate white balance.

Alignment of the right hand side , is a neutral WB.

Both statements can't be true so someone is wrong. Please clarify.

Ken

[Mike Buckley]

Both statements can't be true so someone is wrong. Please clarify.

I'm not positive, but I believe Steve and I are mentioning two different alignments. Even so, I look forward to learning exactly what he means because I'm not getting it.

[DanK]

Ken,

Steve is incorrect. The alignment of the right hand side reflects only the color mix among the brightest pixels in the image. They need not be balanced. For example, imagine an image of a bright red flower, against a dark green and brown background. The right hand side of the histogram would be predominantly red.

I'll give you a concrete example. The first histogram is an image with a whiBal card taking up much of the image, after white balance. The peak is the card. The second histogram is the same image with the whiBal removed, with the same (that is, the correct) white balance applied. After these, I will post the image.







Dan

[Manfred M]

My answer would along the same line as to what Dan and Mike have given. A histogram is a graph showing the distribution (from the darkest value on the left to the lightest on the right) on the x-axis (horizontal axis of the histogram) and the quantity of each specific value recorded on the y-axis (from no pixels on the bottom to maximum possible count at the top of the vertical axis), as recorded in the image. It gives no other information about the image.

Any colour cast would be buried in the overall data across two channels (although in theory it could be across just a single channel). I see no way of disagregating from all the other channel data. If you look at a white balanced image, the shape of each individual channel will reflect this. As an example, if the image has a pure yellow cast, the blue channel should show an increase in the blue values upon colour correction, but the red and green channels would not change at all.

[Mike Buckley]

I'd like to repeat that I look forward to hearing from Steve on this issue again. Judging from his photos and his posts that I have read, my vote would be that he understands this stuff. I don't understand exactly what he means by "alignment of the right hand side" because I don't understand what is being aligned. So, I think responding to that phrase without a clear understanding of what he means is likely to assume something that he really doesn't mean. Just my thought.

Regarding the possibility that the coincidental alignment of the color channels has anything to do with white balance as questioned by the OP, I stand by my and Dan's comment (EDIT: and now Manfred's comment) that it doesn't.

[Steve S]

What is the color mix (R G B ) values for white?

[Mike Buckley]

Steve,

You're being a bit cryptic.

[Steve S]

Steve,

You're being a bit cryptic.

Ok, lots of people use a grey card to set white balance. What is the color mix for 50% grey?

[DanK]

Ok, lots of people use a grey card to set white balance. What is the color mix for 50% grey?

Gray cards used for exposure are not 50%. I believe they are 18%. whiBals are designed to set white balance, not exposure, and they are a bit lighter because they obtained better results in terms of spectral neutrality that way. I think they are around 12%, but I can't find the documentation. To make the cards spectrally neutral, they keep the a and b Lab channels close to 0. It's explained here: http://michaeltapesdesign.com/whibal.html

Lightroom adjusts for WB by setting pixels from the card to even amounts of R, G, and B. I looked at a few of mine, and they were in the range of 72-74, depending on the image.

Note that in the example I gave above, the pixels representing the card were roughly 3/4 of the way from left to right in the histogram. That reflects the amount of illumination on the card. It could be located elsewhere, depending on the image and exposure.

[Steve S]

Gray cards used for exposure are not 50%. I believe they are 18%. whiBals are designed to set white balance, not exposure, and they are a bit lighter because they obtained better results in terms of spectral neutrality that way. I think they are around 12%, but I can't find the documentation. To make the cards spectrally neutral, they keep the a and b Lab channels close to 0. It's explained here: http://michaeltapesdesign.com/whibal.html

Lightroom adjusts for WB by setting pixels from the card to even amounts of R, G, and B. I looked at a few of mine, and they were in the range of 72-74, depending on the image.

Note that in the example I gave above, the pixels representing the card were roughly 3/4 of the way from left to right in the histogram. That reflects the amount of illumination on the card. It could be located elsewhere, depending on the image and exposure.

That doesn't matter, the R G B values are the same, meaning the histograms are aligned. If you have a value of R128 G128 B128 you will see 50% grey.

[Steve S]

In a case where there is no white in the image, it's best to adjust the grey point


Your image...............




With grey point adjusted.................




Adjusted the way i like it, keeping some of the cast so it doesn't look so cold.................

[stuck]

In a case where there is no white in the image, it's best to adjust the grey point

Umm, except from my reading of Dan's post, the flower is supposed to be the yellow colour it is in his post, I read his text as saying the flower's right because he's adjusted it with the aid of his whiBal, which means your 'corrected' versions are not even close to correct.

Or have I misunderstood Dan's post?

Ken

[Steve S]

Colors are tough to get a handle on. Everything i've learned has been self taught, and countless hours learning it. I'm not shure what i do is 100% correct, but i like the results i get, and that's all that matters to me.

[Steve S]

Umm, except from my reading of Dan's post, the flower is supposed to be the yellow colour it is in his post, I read his text as saying the flower's right because he's adjusted it with the aid of his whiBal, which means your 'corrected' versions are not even close to correct.

Or have I misunderstood Dan's post?

Ken

Saturation and luminosity of the colors are not the same as the white balance.

I've had this happen to me shooting yellow as well. (washed out and almost white looking) If you don't shoot under great lighting conditions, the yellow gets washed out. I think the problem is the yellow needs a lower exposure, to make it look yellow (luminosity). The problem is , when you drop the exposure under bad light, the rest of the image looks way underexposed.


Dan's original image has a very distinct red cast to it, and a slight yellow cast. (you can see it in the middle with the muddy looking greens) Which is telling me, the histograms were red to the right, green in the middle, and blue to the left. I would be willing to bet, if dan shot this image under better light, the colors would look correct and the yellows would be saturated.

[DanK]

Umm, except from my reading of Dan's post, the flower is supposed to be the yellow colour it is in his post, I read his text as saying the flower's right because he's adjusted it with the aid of his whiBal, which means your 'corrected' versions are not even close to correct.

Or have I misunderstood Dan's post?

Ken

Nope. You understood. The image I posted is accurate. Because there is nothing approaching a spectrally neutral area in the image--usually the case in images like this--I started with the identical shot, in the identical lighting, with a whiBal card in the frame. I used that area in lightroom to find the color temperature that rendered that area neutral, which is trivially easy in LR. The colors were way off to start, because I shot this under halogen lights, but setting WB this way made the colors very accurate. I then copied that color temperature to the remaining images in the set, creating a correct color balance in all of them. How anyone likes or dislikes the image is immaterial. one can deviate from correct WB at will to get an effect you want. (I often go a tad warmer.) the point of my post was to show the two relevant histograms in an image with correct white balance--one with the spectrally neutral card in the image, the other without it. Those images show clearly that the histograms need not align at the right.

[davidedric]

From another angle (I think!). With most graphical representations, the Area Under the Curve (auc) represents some real physical quantity. Can anyone enlighten me as to what the auc of a histogram represents? As a specific​ example, the difference between a flattish histogram and a higher one, with the same relative peaks heights from left to right?