Why do I get different white balance results from a grey card vs a white card?

Question

I recently purchased a grey card set from here (comes with a grey card, white card, and a black card):

http://www.amazon.com/DGK-Color-Tools-Reference-Correction/dp/B001KNP3MQ/ref=sr_1_1?ie=UTF8&qid=1412290928&sr=8-1&keywords=grey+card

I took a photo that had both the white card and grey card in the image. Then I opened the photo in photoshop and adjusted the WB using the eye droppers inside the 'Levels' dialog. I created two separate layers - layer 1 had the original image with the grey point selected on the grey card, and layer 2 had the white point selected on the white card.

It is my understanding that both layers should produce the same white balanced image, assuming that the image is not overexposed for the white point (which it was not). But the colors on both images were different! Can someone please explain why this is?

Answer 1

Clicking with the white eye dropper does not adjust white balance, but the white point, ie. the upper end of your histogram. do this in the curves dialog and you can easily see the effect on the curves. the grey dropper adjusts the middle points of them.

whatever you click with the white dropper becomes 255,255,255, which is why (you thought) you had to do the telling "75%" adjustment because it seemed too bright to you.

But actually, the "correct" (note that creative considerations are probably more important than technical correctness) adjustments should be achieved by using black, white and grey droppers on their respective patches.

Answer 2

When you click on the droppers in Photoshop, they are assuming that the object you select is neutral RGB, meaning equal values for R, G, and B. If your white balance card is actually color corrected to represent equal values, then the image should color match fairly close to 'real' color (our eyes are easily fooled as to what is white in real life).

However, if the object is not equal R,G,B, there will be a cast. You will see this often if you chose white objects in an image as a stand-in for a white balance card. For example, white clothing is often a terrible choice, as they contain whiteners and other dyes to make them appear white to human eyes, but often they actually have distinct blue or green casts.

I suspect that either the white and grey cards in your kit are not calibrated equal RGB, or perhaps one is and the other is not. Or, often some cards are not neutral under different light conditions, for example under mercury vapor laps. Hopefully your cards came with a calibration reference, showing the actual tested color calibration. If not, you may need a different card.

I recommend the Whitbal card for these reasons. It is calibrated for neutral spectrum, including under lots of conditions. I am not affiliated with them in any way, just a happy customer.

Answer 3

Since in the "white adjusted" picture the white card itself is yellowish (several samples over 10x10 pixel areas consistently gave me average colors richer in red and green than blue), the key must be in your white balancing technique.

Perhaps you're only balancing colors by a single pixel? There is always noise to consider, and if you happened to balance to a bluish pixel then yellowish picture is what you get.

To avoid such effect, you should balance by average color of an area; either your tool lets you do that directly, or you manually blur the area to even out the randomness of single pixels caused by noise.

Answer 4

The white and gray cards are giving you pretty close to the same color balance. The average rectangles from the white and gray cards normalized so that the brightest component is 1 are:

.866, .918, 1.000 white
.846, .941, 1.000 gray

These are actually quite close in terms of human perception, so making either of those a shade of gray would result in pretty much the same color balance.

Here are the pictures ballanced from the white, gray, and white cards next to each other. Note that it is difficult to see a color cast difference.