I guess this should go to the list as well...
···
---------- Forwarded message ----------
Date: Fri, 26 Mar 2004 08:45:59 -0500 (EST)
From: Georg Mischler <[email protected]>
To: John Sutherland <[email protected]>
Subject: Re: [Radiance-general] Candle lighting? rgb values...
John Sutherland wrote:
When you say weighted average for calculating the rgb values, I don't
understand how you mean to calculate that?If my RGB values were
0.759 0.240 0
Do you mean (R+G+B)/3? Or a different calculation? Cheers John S
To determine luminance from RGB radiance values (or the other way
round), you need to take the CIE photometric curve into account.
For that purpose, Radiance uses the following weighting factors:
luminance = 179 * (0.265*R + 0.670*G + 0.065*B)
First of all, you need to make sure that your given color values
average to 1, or you'll end up with skewed results anyway.
Then you take the values that represent weighted unit radiance,
and scale them with the brightness of your source (ie. 17.78).
Weighted color average:
0.502*0.265 + 0.239*0.670 + 0.043*0.065 = 0.296
Weighted unit radiance values:
R: 0.502/0.296 = 1.696
G: 0.239/0.296 = 0.808
B: 0.043/0.296 = 0.145
Source radiance values:
R: 1.696*17.780 = 30.159
R: 0.808*17.780 = 14.358
R: 0.145*17.780 = 2.583
Weighted color average:
0.759*0.265 + 0.240*0.670 + 0.000*0.065 = 0.362
Weighted unit radiance values:
R: 0.759/0.362 = 2.097
G: 0.240/0.362 = 0.663
B: 0.000/0.362 = 0.000
Source radiance values:
R: 2.097*17.780 = 37.296
R: 0.663*17.780 = 11.708
R: 0.000*17.780 = 0.000
Personally, I'd consider colors with a zero component slightly
suspiciuous, but I don't know enough about candle light to tell
whether the ones you found are right or wrong.
The other thing is, that the standard CIE photometric curve
assumes photopic vision. In a scene lit by a single candle you
may end up at least with mesopic vision in reality, which would
theoretically require different factors. I'm not sure how much
different the results would be, though, and whether such a
difference would be significant.
-schorsch
--
Georg Mischler -- simulations developer -- schorsch at schorsch com
+schorsch.com+ -- lighting design tools -- http://www.schorsch.com/