Digital BW, The Print

Paul

I just sent you this off-list as well.

OK here is my attempt to explain luminance scaling with respect to XYZ_Y and
not L*.  It¹s a simple explanation and represents my understanding.  I have
asked a guy who should know to give me his explanation and I¹ll make sure
you hear it.  Here goes:

People read too much into ³Lab is how the eye sees².  My understanding is
that humans perceive luminance first and colour second.  Ok so we need a
luminance component and a hue component.  Lab fits this bill but so does
XYZ.  To quote Bruce Fraser, Lab originated out of an ³attempt to create a
space that is perceptually uniform ­ in other words, distances between
points in the space predict how different the two colours will be to the
human observer.²  L* is ³approximately the cube root of the luminance value
Y (which is a rough approximation of our logarithmic response to
luminance).²  The bit in parentheses is critical.  Again from Bruce ³the
primary Y doubles as the average luminance function of the [eye¹s] cones.²
The eye sees luminance as described by XYZ_Y not L*.  It makes sense then to
scale for white point and black point in XYZ_Y.

It is this Y that we use when talking about density = -log10(XYZ_Y).  Logs
are nice because they turn non-linear responses such as the eye¹s
sensitivity to light into straight lines.


Cheers

Steve

> From: Paul Roark <paul.roark@...>
> Reply-To: <DigitalBlackandWhiteThePrint@yahoogroups.com>
> Date: Tue, 18 Oct 2005 11:52:55 -0700
> To: <DigitalBlackandWhiteThePrint@yahoogroups.com>
> Subject: RE: [Digital BW] ICC v. Transfer Function in Epson driver
> 
>> 
>> ... the ICC requires white point scaling in
>> XYZ for media relativity and Adobe does their black point compensation by
>> scaling in XYZ. 
> 
> Perhaps the "problem" is that XYZ and Lab L have different luminance
> distributions.  The X-Rite "Color Guide" I have gives some very brief
> explanations of these spaces.  With respect to XYZ, it notes, "The basic CIE
> color space is CIE XYZ.  It is based on the visual capabilities of a
> 'Standard Observer,' ..."  On the other hand, in describing L*a*b* it notes
> the "unbalanced nature of the XYZ space ..."  "As a result, the CIE
> developed more uniform color scales called 'CIE L*a8b*' and "CIE L*u*v*.'"
> 
> Now, I think color was the context of the descriptions -- it is called
> "color management" after all -- but I wonder if the same issues affect the
> nature of the grayscale distributions.
> 
>