Michael J. Mahon wrote:
Just thinking about it...the luminance channel is mostly green,
so low luminance should preferentially decrease the proportion of
green...
Shouldn't the *proportion* stay pretty much the same? Of course the
absolute value will indeed drop more.
Also note that I didn't take into account the gamma issue. I made
another program that transformed the resulting colors from SMPTE 170M to
sRGB color space, taking account of gamma; but the result looked worse
on my CRT PC monitor, which probably isn't exactly an sRGB device.
Is it possible that the matrix is a little off?
I derived it from the following by solving for R, G, B:
Y = 0.299 * R + 0.587 * G + 0.114 * B
U = 0.493 * (B - Y)
V = 0.877 * (R - Y)
As far as I know these formulas are correct by definition, but since I
have never read the original standard documents I might be mistaken. In
the derivation I worked with fractions to prevent dropping accuracy
earlier than needed. Maybe I made a mistake, but I'm too lazy to look
for one at this time.
Probably the angles I used are not quite right, as they are adjustable
in a real apple, and might have been adjusted optically to give "good"
yellow and brown in the factory.