White Balance - in the eye of the beholder?

It seems apparent to me that the majority of the landscape, at least the part beyond the lake, is shaded in the first shot and getting direct sun in the second. Therein lies the reason for the colors looking different. On this partly cloudy day, a shaded area will receive a lot of light from the blue sky.

There is no single correct white balance setting for an image that is partly in direct sun and partly in the shade. For instance, if you shoot a snow-covered mountain with the sun coming from the right or left. Set your WB on the sunny side of the mountain and the shaded side will look blue. Conversely, set your WB on the shaded side of the mountain and the sunny side will look yellow-orange.

Curtis N wrote in post #3324363
It seems apparent to me that the majority of the landscape, at least the part beyond the lake, is shaded in the first shot and getting direct sun in the second. Therein lies the reason for the colors looking different. On this partly cloudy day, a shaded area will receive a lot of light from the blue sky.

Sorry Curtis, but I don't see it that way. However, if you swap 'first' and 'second' in your observation, then I think I would agree with you.

In the first shot the sun is pretty much full on to the hill across the lake. You can see my shadow at the bottom of the photo, just right of centre. It's clear that the sun is almost directly behind me, and there are no clouds affecting the main part of the hillside opposite.

In the second photo, you can again see my shadow. This time it's over towards the extreme right. The sun is coming over my left shoulder, and the hillside opposite is shaded by the big cloud which you can see at the top of the picture, just left of centre.

Does that make sense? Have you made a simple typo or am I horribly, utterly confused?

Yeah, you're probably right. You were there and I wasn't, so I'll take your word for it.

Could be the biggest difference between the two images is exposure. The first one is certainly darker, even the blue sky is darker.

But I'm still guessing the differing colors have something to do with the moving clouds, and possibly light reflecting off the lake and back up on the mountain.

Or maybe I'm full of crap.

StewartR wrote in post #3324231
I didn't shoot with a grey card, so I have to fix the WB in PP. I don't really need to nail it, but I'd like to get it "right" (if there is such a thing as "right").

I described my technique/process earlier. In these pictures there aren't any objects that we know to be neutrally coloured, so I took about a dozen samples from the clouds and averaged the results. My concern is that that doesn't seem to be as reliable as I expected it would be.

My guess is the unreliability has to do with the nature of the eye dropper tool itself. If you super zoom all the way in on anything that seems to have a uniform shade/colour when you get down to the pixel level each individual pixel can have a very different colour than its neighbor, but when you zoom out they blend together to be uniform. Maybe this has to do with our sensor photo sites recording only one colour, and the algorithm that takes these values and then makes a colour out of them is the cause? I don't really know, just a guess.

But the way I look at it since our brain does white balancing for us in real life, choosing a photo white balance is nothing more than a matter of taste. I just adjust the temperature slider until it looks the most pleasing to me, I don't even bother myself with thoughts of accuracy anymore because what is accurate is more subjective than quantifiable.

i use a pretty weird way to set my white balance with lightroom.

i'll crank up the vibrance to 100%, then it really brings out the purple or green or blue or orange in your neutrals. then i adjust the tint first and then white balance until it is as neutral as it will get. then crank the vibrance back down. of course it'll only work with pictures that have white or gray in them, but it works.

If I've taken a series of shots in the same, or very similar light, I'll usually play with one of them until I get the WB where it looks "right" to me, then I'll just select all of the other images shot in that same light, and apply that color temp to all of them en masse.

I then go through and adjust things like contrast, saturation, etc., and maybe slightly tweak WB as appropriate for each individual shot.

I have come to like a process described in one of Scott Kelby's books.

• Add a Threshold Adjustment layer and the find the points on the image where the darkest/lightest pixels begin to poke through and anchor your sample points there.
  

• Discard the Adjustment Layer
  

• Add a new Layer with Blend mode = Overlay, fill with 50% Grey
  

• Change Blend mode to Difference
  

• Add Threshold Adjustment Layer and place slider all the way to the left (darkest).
  
• Move slider to the right until the first pixels begin to poke through which is where you place the 3rd sample point. This will be your middle tone area.
  

• Discard the Threshold and Difference layers
  

• Add a Curves adjustment layer. Use black eyedropper (use Caps lock for a different icon than the default one) to click on the 1st sample point, the white for #2, and midtone for #3.
  

HOSTED PHOTO
please log in to view hosted photos in full size.

ayotnoms -> thanks for sharing! I must try that out sometime.

Not that this'll help much, but I did a "Color Match" with CS 2, matching the second with the first. Much of the color came out very similar, but the second did have a much rosier hue on the hill on the other side of the lake; a little to pinkish I'd say. I then tried the white balance in curves (using the middle gray tones) and got a HUGE variety of different looks. The only way to know which one is the best would have been to actually be there. Like you said, they're all satisfying, but which is "real"?

Thanks for all the suggestions, everybody.

ayotnoms, that looks like a really good technique. However I think it implictly assumes that the brightest pixel in the picture is supposed to be white and the darkest one is supposed to be black. That might not always be the case. (Though I guess more often than not it would be pretty close.)

I'm still a complete novice with CS2 but I'm more confident with PS Elements. However that technique uses tools that don't exist in PSE. So this is what I did instead to each image:

• I assumed that on average the clouds should be neutral (white/grey).
  
• I selected the densest areas of the clouds using the 'magic wand' tool and copied them to a new layer. (Only the densest areas, to avoid contamination from the blue sky.)
  
• I examined the colour channel histograms as in the earlier post.
  
• I used the levels controls in each colour channel to bring the brightest levels of each channel up to the same value. (For example, in the histogram above, the green channel needs to come up a bit and the blue channel needs to come up more.)
  
• This gave me clouds which were neutral in colour. I noted the adjustments that I had made to each colour channel.
  
• I deleted the clouds layer and applied the same levels adjustments to the whole image.
  

Byte size: ZERO | Content warning: NOT AN IMAGE

Byte size: ZERO | Content warning: NOT AN IMAGE

StewartR wrote in post #3336255
... However I think it implictly assumes that the brightest pixel in the picture is supposed to be white and the darkest one is supposed to be black. ...

Actually as it pertains to your--very nice--image, it explicitly reveals the pixels* closest to white/black as you can see in info palette : #1 = R(13)G(11)B(11) #2 = R(253)G(254)B(253). The threshold method is a matter of black or white. If you happen upon the darkest pixel and find that it isn't black it will be because the values between the 3 channels are just a bit off...again, as the values in the Info palette show. The technique is built on leveraging Photoshop's ability to objectively identify color values.

No Big Deal when it's all over with, as you have succinctly stated...

StewartR wrote in post #3336255
What matters is that I'm happy.

Cheers!
* I say pixels because the eyedropper tool option was set to 3 X 3 average

I almost always balance to make it look good, then do a test print as final check. One can get very technical and have a poor image - colour can be very subjective.

I'm convinced that most sunsets as submitted have been "re-balanced" to get them warmer - either that or I've never lived where the sunsets were any good.

ayotnoms wrote in post #3337872
Actually as it pertains to your--very nice--image, it explicitly reveals the pixels* closest to white/black as you can see in info palette : #1 = R(13)G(11)B(11) #2 = R(253)G(254)B(253). The threshold method is a matter of black or white. If you happen upon the darkest pixel and find that it isn't black it will be because the values between the 3 channels are just a bit off...again, as the values in the Info palette show. The technique is built on leveraging Photoshop's ability to objectively identify color values.

Thanks Steve. I think I need to (a) wrap a cold towel around my head; and (b) put a bit more effort into learning what Photoshop (as opposed to PSE) can do. I appreciate the effort you've put into explaining the concept to me.

StewartR wrote in post #3338428
Thanks Steve. I think I need to (a) wrap a cold towel around my head; and (b) put a bit more effort into learning what Photoshop (as opposed to PSE) can do. I appreciate the effort you've put into explaining the concept to me.

Truth is, my philosophy about color correction is closer to your "What matters is I'm happy" position. All this technical stuff is B O R I N G