Apple

Hi Roger,

In a message dated 9/14/2005 12:51:10 P.M. Pacific Daylight Time, email@hidden writes:

<<And to really implement CIECAM requires a *lot* of measurements beyond mere CIE colorimetry. It's not going to be for the masses IMO. Unless MS does all kinds of simplifying assumptions, which could be debated. >>

I've evaluated several appearance models for digital photography. Color appearance models as we commonly think of them are merely CIE colorimetry plus viewing conditions. I think the measurements you are referring to are primarily viewing conditions. For DP, this is readily available (or estimable) from the image and image meta data, albeit imperfectly. Put that together with some assumptions about output viewing conditions (a la ICC) and you can do pretty well. CIECAM has attractive attributes such as chromatic adaptation (ability to interchange amongst illuminants), colorfulness (ability to interchange amongst different illumination levels), and improved hue and chroma scales (useful for large color moves like gamut mapping). On the downside, it can be numerically unstable near the gamut boundary when input and output viewing conditions are quite different. And no Helmholtz-Kohlrausch compensation. Also the luminance function leaves something to be desired.

The appearance-based approach combined with a smart CMM offers a great deal of potential, and CIECAM02 is a reasonable place to start. There are simpler and better appearance models on the horizon, as yet unpublished. It appears WCS contemplates this inevitability and will be extensible. This approach comes at the cost of considerable complication so it remains to be seen how well the implementation shakes out. But even as an avowed Apple snob, I have to say kudos to MS for forward thinking.