Thank you Graeme for your reply,
I missed the "gamma" documentation on argyllcms -- gee! -- where you outline
the various known approaches to calculating gamma, using least squares and so
on. Thank you for those invaluable explanations. My bad for not finding them
earlier.
I won't take much of your time simply to comment that, using Excel Solver,
based on the Luminance measurements collected on my NEC display, in "Gamma 2.2"
mode, I am getting an estimated value of "2.16" (!) which I find quite
honorable. (Had to brag about it...)
So, in essence, given this superlative gamma performance, I guess I would not
have to get into a LUT-base profile?
In your reply, below, you wrote: "...generally the default will work".
According to the online documentation, Colprof default is :
-a lxXYgsmGS Algorithm type override
l = Lab cLUT (def.), <---------------------------------
x = XYZ cLUT,
X = display XYZ cLUT + matrix,
Y = display XYZ cLUT + debug matrix
g = gamma+matrix, s = shaper+matrix,
m = matrix only,
1) Does that mean that, the default is to build a Lab cLUT?
2) Only if the '-a' parameter isn't specified, right?
3) Does colorprof needs a minimum of measurements to build a cLUT profile?
3a) Will I get a warning or error if I have, say, only 20 measurements?
I'm sorry to put you through all these questions ☹
/ Roger
-----Original Message-----
From: argyllcms-bounce@xxxxxxxxxxxxx <argyllcms-bounce@xxxxxxxxxxxxx> On Behalf
Of Graeme Gill
Sent: Saturday, February 1, 2020 2:13 AM
To: argyllcms@xxxxxxxxxxxxx
Subject: [argyllcms] Re: Display profile options
graxx@xxxxxxxxxxxx wrote:
Hi,
generally the default will work. If you have a wide gamut display, then
perhaps you would need XYZ cLUT.
(colprof will warn you if your display primaries are out of gamut of the ICC
L*a*b* encoding.)
If you know your display is will behaved (i.e. additive) then shaper+matrix may
give a smoother result. One way of deciding is whether the self fit errors are
OK for the matrix.
If you want to use a cLUT but some of your applications can only use a matrix,
then display XYZ cLUT + matrix may be the best option.
If you want to figure out which software is using the matrix rather than the
cLUT, then display XYZ cLUT + debug matrix is the right option.
If you know that your display has an exact power curve characteristic, then
maybe gamma+matrix would give a smoother result. (Again, self fit error is a
hint.)
Everything else is special cases for applications that are particularly fussy.
For instance at at least one stage Photoshop would only handle a matrix profile
with a single gamma curve as its display profile.
Matrix only is for raw camera profiles.
But I am told that this model is not "invertible"? IOW, I'd have no
way to modeling the XYZ to RGB direction?
