Re: [png-mng-implement] Re: theory of gamuts, primaries

[John D. <js...@av...>]

Alas I wrote in haste:
 >> I don't think anybody would perceive [X,Y,Z] = [1,0,1]
 >>as being equally bright as [0,0,0], even though they have the
 >>same "middle tristimulus value".

On 03/19/05 22:27, John Bowler responded:
> 
> Transform it [1,0,1] back to D65 RGB and take a close look at the values
> of the components.  
...
> Not all the values in XYZ space correspond to real spectra.

Touché ... I used a bad example.  But the larger point I was
making remains valid, as I can demonstrate by taking up JB's
suggestion.  Here are some XYZ values at constant Y (!) and
the correspponding sRGB values:

     X       Y       Z          R       G       B
   0.039   0.077   0.014      0.000   0.107   0.001
   0.039   0.077   0.015     -0.000   0.107   0.002
   0.039   0.077   0.018     -0.000   0.107   0.005
   0.041   0.077   0.026      0.000   0.106   0.014
   0.044   0.077   0.049      0.000   0.104   0.038
   0.053   0.077   0.104      0.000   0.098   0.098
   0.072   0.077   0.228      0.000   0.084   0.229
   0.104   0.077   0.440     -0.000   0.062   0.455
   0.141   0.077   0.682     -0.000   0.036   0.714
   0.169   0.077   0.859      0.000   0.017   0.902
   0.183   0.077   0.950      0.000   0.007   0.999

Now it's true that the eye is more sensitive to green
wavelengths than it is to extreme short or extreme long
wavelengths, but the sRGB phosphors are not so extreme
as to create a 9-to-1 difference, as the numbers above
would seem to allege.

And FWIW, in an abundance of thoroughness, I fired up
ghostscript and compared
    0.000   0.107   0.001 setrgbcolor
against
    0.000   0.007   0.999 setrgbcolor
and was totally unsurprised to observe that the latter
looked a lot brighter.

So tell me, what reason is there to believe that the Y
value from the CIE XYZ tristimulus vector corresponds
to perceived brightness?