Picking the wrong reference for colorchecker analysis — an extreme example

This is the 20th in a series of posts on color reproduction. The series starts here.

X-rite publishes ColorChecker reference color values by illuminating their targets with D50 light, not by publishing the spectral responses for each patch. If you want to convert the X-Rite values to a color space with a white point that’s different from the xy values that D50 resolves to given your standard observer, you must use an adaptation algorithm like Bradford or von Kreis to get to that color space.

What if you’re lighting your target with fluorescent light, because you’ll be shooting in that light?

My claim is that your reference ought to be what the target looks like in fluorescent light, not what the target looks like in D50 adapted to fluorescent light. We’ve been looking at the errors for smaller lighting mismatches. What happens with a big one?

I took the CIE F4 Illuminant spectrum, and, through the magic of computer simulation, lit a Macbeth target and measured the patches with the white point of F4 as the reference. Then I adapted the values of the target illuminated by D50  to the F4 white point using Bradford and looked at the differences.

The reference and target spectra are quite different.

And so are the results.

If we saw errors like that in a normal profile, we’d say the profile isn’t very good. But these errors are with a perfect (Luther-Ives) camera and a perfect profile. They are the result of picking difference reference illuminations and come from a combination of adaptation errors and illuminant metameric errors.

By the way, you will note that the gray axis errors are small.