2006.12.07 01:28 "Re: [Tiff] Re: Tiff Digest, Vol 31, Issue 3", by Graeme Gill

We don't really dissagree. I personally use the second option I enumerated, i.e. converting RGB to CIE Y and back to greyscale. That seems to be exactly what you propose. Is this an approximation, or is it 'actually correct'? I prefer to avoid these terms. Some people are bound to say CIE XYZ sucks and CIE L*a*b* is about as different from human vision as you can get. They may be right, they may be wrong.

Sorry, I don't buy this "everybody's got an opinion" stuff when it comes to things we can measure. You can quibble all you like about whether the CIE observer curves are the best possible, but the fact is that noone has come up with any alternatives that are convincingly better in matching the measurements in over 70 years. So Given an additive device like an RGB display, we can prove that the CIE Y value is going to correspond as accurately as we know how, to the equivalent monochrome intensity of its RGB output.

Graeme Gill.