Digital Light & Color

A recent experience with white balance adjustment and color cast correction has me puzzled. Perhaps someone can explain what is going on. It involves functions in both the raw converter and in the transformation Color>Balance. It all arises from my experience with some portraits I took recently of the board members of an organization for display on its website and, for the CEO and President, framed in the lobby.

Before describing the situation, let me confirm that immediately before processing the images, I color calibrated my monitor with a device from Pantone (now X-Rite). When shooting the images, I also employed a white balance card from Whibal.

The portraits were taken in a setup using a studio strobe bounced from a white umbrella. All subjects sat on the same stool so the lighting was identical for all images. The first image was taken with the subject holding the Whibal card.
In the Color tab of PWP’s raw converter, I used the picker to probe the Whibal card in the first image. It indicated a color temperature substantially different from the camera white balance. I probed the Whibal card both before and after making preferential adjustments to exposure and contrast. Although the temperature result of these probes differed slightly from one another, they were in a relatively narrow range but quite different from the camera- determined color temp.

Following color temperature adjustments dictated by the probe of the Whibal card, I thought the colors (particularly the flesh tones) looked pretty good. But when the processed images were printed and compared to portraits I took last year in the same lighting setup, I realized they were a “little off”. They were just a hair too red (magenta). The subjects had a nice healthy glow -- a bit like they had just come in from a day on the beach in the summer. I confirmed that this did not constitute a variance between my computer display and the printed output. I compared the prints to my computer screen and they looked essentially the same. It was just that I had been insensitive to the slight magenta color cast during processing.

So, to further correct the color balance, I took the fully edited files into the transformation Color>Balance. There, I began by conducting a “reset” and then clicked on the Whibal portion of the image that included the Whibal card. This resulted in no change to the color profile -– the red, green and blue lines lined up perfectly along the 45 degree line. I interpreted this as confirmation that the response of the color balance transformation was identical to the color picker within the raw converter, because that was the source of the color balance I used in the edited images.

Then, in the color balance transformation, I zoomed way in and clicked on the white of the subject’s eye. There was an immediate change in the red, green and blue settings. More importantly, the redness of the flesh tones disappeared and they looked “perfectly Caucasian for November in the northern US.” The resulting prints were equally well color balanced.

So my questions are as follows:
1. Why did the whites of the eyes provide a better color balance adjustment than the expensive Whibal card?
2. I would like to be able to probe something a small as the whites of the eyes in the raw converter color tab. But the raw converter does not permit zooming in to 1:1 so it may be difficult insure that the picker is really selecting only this tiny area. Why is the zoom range in the raw converter limited?
3. Finding that probing something pure white produced a better color balance adjustment than probing the gray Whibal card was contrary to anything I have previously read. Can anyone explain why that would be? Perhaps I’ll switch to using a white test patch area in some images. Does someone have a suggestion for a good homemade white media that would duplicate the whites of a person’s eyes?

Doug,

Can it be that light reflected from a coloured surface has fallen on your grey card ?
The eyes are quiet well "shielded" in their sockets, a grey card held in the wrong angle can easily get some "coloured" light.
Outside it can easily happen to hold the grey card so it gets blue light from the sky and so ruins the white balance.
Another possibilty could be that you have overexposed the grey card.

Thanks for the suggestions, Dieter.
I don't think there was significant ambient or reflected light from a source other than my studio light bounced from a white umbrella. The ceilings in the room where I was set up are about 10-12 feet high (and the studio light and umbrella were on a stand, aimed downward toward the subject seated on a stool and not upward). The most adjacent walls were at least 10 feet away and were mostly large glass windows. It was dark outside, so no daylight would be coming into the scene.

Overexposing the gray card could be a possibility however. I made sure my settings produced no highlight clipping. But other than that, the exposure could have been mildly ETTR since I know that collects better data and I can always scale back exposure in post. I simply took a test exposure of the entire scene (not the gray card by itself) and made sure the histogram occupied the area right up to the right-hand side, without clipping.

I was not previously aware that perfect middle-gray exposure of the card was critical.
So the issue related to this would be the question of whether slight overexposure of a test card would result in the temperature being skewed to the "warm" side.

Doug,

I do not have expieriencew with the Whibal you mentioned, I took a look on the net and it looked quiet bright compared to a 18% standard grey card I use.
The documentation i found in their website didn't tell either.
A 50% grey in PWP comes very close to the tone of my grey card.
And on a "correct" exposed image the grey card should come out as 50% grey (Zone 5 in the Zone system).
I never did tests what happens if i radically overexpose the card, one stop does not hurt, in my expierience.

Check the red green and blue histograms of the image with the Whibal target to make sure there is no clipping. ETTR can easily lead to some clipping if large spikes in the histogram scale lower values into invisibility. Try usingl histogram expansion when checking histograms for ETTR.

Thanks for weighing in, Jonathan.

This remains a mystery to me. Based on Dieter Mayr's conjecture that it might be related to overexposure of the Whibal white balance target, I subsequently established contact with a person (not a PWP user) who has published treatises promoting ETTR. The question was whether the ETTR process can adversely affect white balance determination. I shared the raw files with him and, using other software, he determined that in the Whibal target, the green channel was 1 EV below clipping, the red channel 2EV and the blue channel 1 1/3 EV below. So while exposure of the target was certainly above "middle gray", it wasn't clipped.

He also processed the raw files in different software (he mentioned RPP and PhotoNinja) and claimed to hot have experiences the too-red color correction that I did with PWP.

Since my original post, I have converted from raw the test photo using the camera white balance temperature (which is 5129 K) and then taken the resulting tiff to the transformation Color>Balance. When probing the test patch area in that transformation, I get an adjusted color balance that looks very much like the result when probing in the raw converter (although no temperature is provided in the Color>Balance transform). When the test patch is probed in the raw converter it says the temperature is between 5900 and 6000 K).

Notably, when I probe the whites of the eyes in the Color>Balance transformation, I approach what I think is a more natural looking skin tone. Not quite so sun-tanned -- which none of us are in Ohio in November. LOL

If you would like to look for yourself, the files are at Dropbox location: https://www.dropbox.com/sh/kh3ywqet3fld ... U9dSa?dl=0

The file DSC_0005 is the raw file containing the Whibal test target.
The file DSC_0038 is the raw file with the portrait of one of the subjects in the same lighting setup.
The file titled "Stang" (last name of the subject) is the JPEG from the fully edited (cropped, shadows lightened, sharpened, etc.) result from DSC_0038. This version does NOT reflect a white balance based on the Whibal target -- it is the result of fiddling in the Color>Balance transformation.

Doug, I have played around a little with your pictures.
DSC_0005.NEF: When probing the white field of the card gives a very warm (red) result - about 6300K color temp.
Probing the grey part of the card or the grey background results in 6000K color temp, and a good result for my taste.
Probing the white of the eyes gives much too cold results for me, something around 5300K with a (for me) ugly cyanish tint.

DSC_0038.NEF:
I get a acceptable result when probing the inner part of his right eyes white, his left eye goves to cool results as above.
Probing the grey background gives the same 6000K as above, and a good result for my taste, too.

So I would conclude that probing the grey part of the card gives correct results, but on portraits I would never trust "correct" results, but season to taste, based on that correct result.
Personally, mostly I like portraits to be a bit warmer then "correct", even a little too cold is a no go for me.

Thanks, Dieter, for doing some work on this ....
I'm glad you read a white balance of about 6000 which confirms that I'm not somehow getting a different result. It is just that for my taste, I thought it produced an overly warm result -- unless the subject had just spent the afternoon at the beach.

Did you look at my final JPEG of the finished headshot of one of the subjects? That was not processed at 6000K (although that's where I started). I scaled back the "warmth" significantly through "eyeball" adjustments of the different color channels in the Color>Balance transformation. It is my version of your admonition to "season to taste". Did you think it was too cold or cyanish? (Does PWP provide a way to measure the temperature after that "eyeball" adjustment to see how it compares to 6000?)

The Whibal card is clearly exposed to more than "middle gray". So are you concluding that higher-than-mid-range exposure of the target (so long as it is not clipped), is not causing an incorrect white balance reading?

Doug

Doug,

I took your Jpeg picture and even though it looks kind of redish, it has actually a blue tint.
I have developed your NEF with 6000K and measured the grey background with the readout tool.
Neat the Top right corner I measure quiet exact neutral grey, 30% in each channel, a variation of about +/- 1 or 2 % over the entire background.
Your Jpeg measures to 38 - 38 - 44 RGB %, so a blue tint.
Using Filter transformation I filtered out the blue tint with CC05Y plus CC075Y, giving a summary of CC12.5Y.
The Skin tones compare to them of the developed NEF, too (I didn't adjust the contrast and exposure to match with your JPEG, but the bright areas of the skin give a good comparison).
Then I used Filter on your Jpeg with Filter Color - Color Temperature and could filter out the tint with a value of 5000K.
I'm not into the mathematics of all this, have read something about it long time ago, but not much is present anymore.
I just remember those color temperature calculations are not trivial.

Doug,

For me, it worked well with measuring the grey part of the card, so it should be ok.