Gamma of CIE RGB 1931 color space
Roger Breton
time anyone of you guys elect to reply) but I having difficulty with the
choice of gamma to be associated with the 1931 CIE RGB matching stimuli of
wavelength 700.0, 546.1 and 435.8 in color space transformations.
AdobeRGB, sRGB, ColorMatchRGB and manyOtherRGBs have well documented
chromaticity co-ordinates and reference whites but not the 1931 CIE RGB
matching stimuli. I searched Wijecki and Stiles, CIE Publications 15.2:1986,
Hunt, Berns and a slew of other authoritative texts to no avail.
So, if anyone has any historical background or other conceptual
justification for the usage of gamma 2.2 to be used in conjunction with the
1931 CIE RGB "color space", please come forward and share your knowledge.
I'm going in circles...
Roger Breton
Gernot Hoffmann
Roger Breton schrieb:
Roger,
in W & S, chapter 3.3.3 , here p.136, you'll find:
the units of the primaries are chosen so that
light with equal amounts of each primary is Equal
Energy White. The magic numbers for R,G,B in
the top diagram on p.6 here are based on W&S:
http://www.fho-emden.de/~hoffmann/ciexyz29082000.pdf
The ICC profile CieRGB uses the same primaries
and E.E. whitepoint X=Y=Z=1.0 . This can be
shown by ICC Profile-Inspector (with a tiny rounding
error in Z). Gamma is 2.2 , but that's IMO not the
least related to CIE colorimetry because this is
linear. It's IMO just convenient for the definition
of an artificial working space (like sRGB, Adobe-
RGB(98).
Best regards --Gernot Hoffmann
Roger Breton
> in W & S, chapter 3.3.3 , here p.136, you'll find:
> the units of the primaries are chosen so that
> light with equal amounts of each primary is Equal
> Energy White. The magic numbers for R,G,B in
> the top diagram on p.6 here are based on W&S:
> http://www.fho-emden.de/~hoffmann/ciexyz29082000.pdf
Thank's. Das ist sehr klaar.
> The ICC profile CieRGB uses the same primaries
> and E.E. whitepoint X=Y=Z=1.0 .
Where does this "CieRGB" profile came from? Certainly not from CIE!
> It's IMO just convenient for the definition
> of an artificial working space (like sRGB, Adobe-
> RGB(98).
So you think whoever created the profile chose Gamma = 2.2 out of
convenience?
> Best regards --Gernot Hoffmann
MfG,
Roger Breton
Roger Breton
Adobe.
--
I wonder how to interpret graphically the X = Y = Z = 1.000 tristimulus
values for the Equi-Energy White point?
Roger Breton
Gernot Hoffmann
Roger,
indeed, Adobe 2000/8/11. Perhaps an unlucky attempt
to define a wide gamut working space. Unlucky, because
the red and blue primaries are almost invisible.
With Gamma=1 this ICC profile could be useful for
colorimetric calculations. With Gamma=2.2 it's useless,
IMO.
X=Y=Z means all grays with E.E. white point. It's the line
in XYZ with equal coordinates. X=Y=Z=1.0 is the white
point after the normalization for Y=1.0 (if physical values
in cd/m2 are not available).
Perhaps somebody knows an answer to this question:
Starting color matching experiments, CMFs are not yet
available. How can the primaries in the experiment be
adjusted for Equal Energy ?
An E.E. illuminant (flat spectrum) delivers equal values
X=Y=Z.
This says: areas under the CMFs x-bar, y-bar, z-bar are
equal. Why are the areas equal ?
Best regards --Gernot Hoffmann
Tom Lianza
The normalization is arbitrary and after the fact. The primaries for
the color matching functions do not not need to be normalized for Equal
energy, in fact, they can't be because the equal energy spectra is
flat(as you said). This scaling sets the .333x .333 y point in the xy
diagram.
The primaries don't even have to be the same primaries as used in the
original experiment. For different primaries, the resultant color
matching functions should be a linear transformation away from CMF's
calculated with a different set of primaries. The x-bar, y=bar, z=bar
functions are a set of non-negative CMF's which are linear derived from
the experimentally derived CMF's. If I scaled the negative lobe CMF's
derived from the physical experiment, for equal energy response, they
would behave precisely as the the standard published cmf's for the
measurement of color.
regards,
Tom
Roger Breton
> indeed, Adobe 2000/8/11.
Yes, it was in the ICC profile copyright tag.
> Perhaps an unlucky attempt
> to define a wide gamut working space.
Well, I never, never wrote any criticism about it anywhere on the net! But
it does not mean that it is any good.
> Unlucky, because
> the red and blue primaries are almost invisible.
Invisible? As in can't be displayed on any real monitors? Because the
chromaticity of the red and blue primaries are out of gamut of any present
day monitors? That's how I understand your point.
> With Gamma=1 this ICC profile could be useful for
> colorimetric calculations.
Ahh! I think this is a very interesting observation. As a matter of fact,
how did Adobe ever decided to use a gamma=2.2? Probably out of convenience,
as we said before. BUT, you know, intuitively, I think Wiskecki would argue
that its "native" value has to be 1.0, by definition, since XYZ is a linear
space!!!!!
> With Gamma=2.2 it's useless, IMO.
I think it is even worse than that : it is plain wrong. Conceptually wrong
and contrary to the spirit of the Standard Observer.
I'm sure Gamma=1.0 would please Timo and royally upset some people at Adobe
-- you know who you are...
> X=Y=Z means all grays with E.E. white point.
OK.
> It's the line
> in XYZ with equal coordinates. X=Y=Z=1.0
Can you show that line on any of your graphs?
> Perhaps somebody knows an answer to this question:
>
> Starting color matching experiments, CMFs are not yet
> available. How can the primaries in the experiment be
> adjusted for Equal Energy ?
I think this adjustment cannot be done a priori, Gernot. I believe it can
only be determined empirically.
> An E.E. illuminant (flat spectrum) delivers equal values
> X=Y=Z.
> This says: areas under the CMFs x-bar, y-bar, z-bar are
> equal. Why are the areas equal ?
Yes, it means the area under each curves of the CMFs x-bar, y-bar and z-bar
are all = 1.0 as well. Easy to see by summation (integral) of x-bar, y-bar
or z-bar because they all give the same results.
> Best regards --Gernot Hoffmann
MfG / Roger Breton
Roger Breton
> it does not mean that it is any good.
I meant to write: I never saw any 'written' criticisms of Adobe's preference
of gamma=2.2 anywhere on the net.
Sorry about that. / Roger Breton
Graeme Gill
>> Unlucky, because
>>the red and blue primaries are almost invisible.
>
> Invisible? As in can't be displayed on any real monitors? Because the
> chromaticity of the red and blue primaries are out of gamut of any present
> day monitors? That's how I understand your point.
It's easy enough to define primaries on the spectrum locus
that are at such extreme wavelengths that our eye's have
very low sensitivity to them, and (theoretically) you then
have to drive them at extremely high energies to reproduce
colors at a reasonable brightness.
In practice you would not want to use such primaries,
because it would be highly inefficient and dangerous
(ie. might well damage your eyes), while not actually
increasing the gamut very much.
You could also pick imaginary primaries that lie outside
the spectrum locus, and they are definitely invisible!
(This is what the XYZ co-ordinate system is anyway.)
Graeme Gill.
Roger Breton
> It's easy enough to define primaries on the spectrum locus
> that are at such extreme wavelengths that our eye's have
> very low sensitivity to them, and (theoretically) you then
> have to drive them at extremely high energies to reproduce
> colors at a reasonable brightness.
I read that 700nm is extreme but you can't say that 543.9nm is extreme?
> In practice you would not want to use such primaries,
Of course not.
> because it would be highly inefficient and dangerous
> (ie. might well damage your eyes), while not actually
> increasing the gamut very much.
OK.
> You could also pick imaginary primaries that lie outside
> the spectrum locus, and they are definitely invisible!
> (This is what the XYZ co-ordinate system is anyway.)
Yes. I could do that as well. But, I'm not after the choice of some set of
primaries to synthesize some RGB working space of my own.
> Graeme Gill.
Roger