Shai,
I have also encountered this problem, and the best solution I have found is
semi-empirical. That shouldn't be surprising, since these color models are
somewhat subjective anyway-- your Cyan ink may not be exactly the same shade
as my Cyan, and even if we agree on a standard we might use different
screens so that it gives a different shade of green when screened with a
particular Yellow.? The same CMYK can look significantly different with
different print jobs.? So I established a local standard for RGB-- a scanner
with invariant settings.? Then for a particular set of print conditions, I
print a set of samples: for the linear case it is 5 swatches consisting of
the unprinted paper and one each of pure C,M,Y, and K; for the quadratic
model 15 swatches consisting of the linear ones plus a 50% print of each
primary and a 50%:50% swatch of each possible combination of 2 different
colors; the cubic and quarternary models require 35 and 70 sample swatches
respectively, and the choice of color composition is analogous.
Then you scan them on the local standard scanner.? You now have (in the
quadratic case) 15 sets of defined points ((RGB),(CMYK)), which give 15
equations of the form
R = q0 + q1*C + q2*M + q3*Y + q4*K + q11*C^2 + q12*C*M + [rest of q terms]
containing 15 undetermined q coefficients.? Solve this system of eqns for
the q's.? The G and B cases are analogous; you now have a general mapping
from CMYK to RGB for the particular print conditions.? Notice that you can
make this model as precise as you want by increasing the order of the
solution.
HTH,
Hiram.
?
?
> To all color experts
> I am using the (known) simple (though non-linear) formulas for display
> of CMYK images:
> ==============================================
> CMYK -> RGB
> Red? =(1-Black)*(1-Cyan)
> Green=(1-Black)*(1-Magenta)
> Blue =(1-Black)*(1-Yellow)
> C, M, Y, K, R, G, and B have a range of [0;1].
> ==============================================
> I obtained the above formulas from various faq's and tiff library
> codes.
> The display looks o.k. but is far from the result obtained using
> PhotoShop which looks MUCH better!!! I was told that PhotoShop does not
> use the above formulas but uses instead hard-coded conversion tables (+
> interpolations between table values). It looks like I'm missing a lot of
> know-how (or quite a few hours of reversed engineering) !?
> ??? thanks
> ??????? Shai
?
by