Thursday, November 4, 2021

If CIECAM is the answer, what was the question?

 (Send contributions to )


Imagine, while studying in preparation for his next life as a color scientist, the ghost of Alex Trebek visits us in his former role and announces his truly Final Jeopardy answer:



The contestants blink and Trebek explains:

“If CIECAM is the answer, what was the question?”

And the contestants answer:


Contestant 1: “What model predicts symmetric color matches?” WRONG: That was CIEXYZ.

Contestant 2: “What model predicts asymmetric color matches?” WRONG: That was CIECAT.

Contestant 3: “What model predicts color difference?” WRONG: That was CIECAM-UCS.

Contestant 4: “What model allows a stimulus, in given viewing conditions, to be numerically described with correlates of perceptual attributes such as brightness, lightness, colorfulness, chroma, and hue?” [1] CORRECT: Although CIE’s color-appearance models, CIECAMs, are not the only possible models.


Contestant 2: “That’s not fair! I’ve seen CIECAMs tested by asymmetric matches, but never by the elusive ‘numerically described perceptual attributes.’”


Contestant 3: “Well, come to think of it, Luo et al. [2] describe experiments to test people’s ability to use particular perceptual attributes: ‘For the memory matching method, observers are first trained using the Munsell colour order system (or some other suitable system) until they are very familiar with these scales (i.e., Munsell Value, Chroma, and Hue) … In the magnitude estimation method, observers are asked to make estimates of the magnitudes of some perceptual attributes (e.g., lightness, colourfulness, and hue). It is essential that each observer clearly understands the perceptual attributes being scaled.’”


Contestant 2: “It sounds as if those experiments tested the memorability and amenability for scaling of particular coordinates of a particular color-order system. They cannot make a statement about color appearance independent of the color-order coordinates chosen for training the subjects. How do you know one CAM is better than another if the subject’s training has such a bias? And I understand the precision of these tests is pretty low. I still think there is no match-free way to test a CAM—or for that matter, to use a CAM for color management. Alex is wrong and we should have a recount.”


Trebek: Well, it’s time for me to go now. This discussion is turning into a quagmire, and it looks like real color-management systems rely on asymmetric match predictions anyway. So let’s ask a professional organization like the ISCC to sort it out. Meanwhile, I’ll have to tell my game-show successor that the right question for CIECAM is “What color-management model is not out of Jeopardy?”


[1] M. D. Fairchild and L. Reniff, A pictorial review of color appearance models, 1997 SID/IS&T Color Imaging Conference, first paragraph of Introduction.

[2] M. R. Luo et al., Quantifying colour appearance part I. LUTCHI colour appearance data. Color Res Appl 16: 168-180 (1991).


Michael H. Brill