[argyllcms] Re: Calibration + profiling or Profiling only

  • From: Florian Höch <lists+argyllcms@xxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Sun, 12 Sep 2010 01:34:50 +0200

On the other hand, madVR supports full 3DLUTs. It is theoretically
better than a little ICC file. (Of course, it also depends on the
creator software. But I follow the development of the yCMS software and
I made some tests. I like it.)

I've looked briefly at the 3DLUT format and it's capabilities seem similar to a shaper/matrix ICC profile (basically such a 3DLUT is a file precalculated and interpolated from measured RGB primaries, whitepoint and IRE at several levels - so the term LUT has not the same meaning as for an LUT ICC profile, where it can be a look up table calculated from an virtually unlimited amount of actual measured values at different points in device space, not just primaries/whitepoint and some grayscale). This means it won't be able to accurately characterize a nonlinear/not very additive device (atleast color-wise), like most consumer LCD displays today, which a LUT ICC profile can do. For movie viewing, where color accuracy is probably not as important, it could still serve its purpose well I think.

 >>> I'm not sure I understand what you mean by 'plain tonal response'.
Can you elaborate further.

A strange curve which doesn't really follow an exponential curve but
averages around an exponential curve with a low exponent (~1.73).

Got it. Thanks for the explanation.

 >>> All three aim towards the same target, so I think it's only a
difference how much 'work' each of the three does.

Ok, let me redraw this...

The white POINT can be corrected with the hardware RGB gains. It will
permanently reduce the contrast ratio and the number of the available
shades (the amount of this is unknown now but I guess it's good because
it is done by a 12 bit controller).
Now, the VGA LUT will correct the white balance across the full gray
ramp. Because the curves are not fully coherent, there is a chance that
it will decrease the R and G at the middle of the grayscale because I
decreased the B to get a better WP at the 100% IRE.
Of course, it shouldn't be a big thing, but I am talking about the
accumulation of these little things when there is numerous correction
steps. Do you get the idea...?

I just don't think it will sum up to anything that will even be close to noticeable. I wouldn't worry about it imo.

Now, let's continue with the gamut emulation.
It cannot be done by the VGA LUT. But the VGA LUT will correct the gamma
curve as well. So,
I think the gamut emulation is tricky and it mess with the WB and gamma too.
Let's see an example. (But tell me if I am wrong. I am not sure.)
You have an oversaturated green, so you have to add some red and blue.
If you do so, you will increase the luminance. So, you have to decrease
all of the three values to get the same chroma with correct luma.
(Of course, in the real world, you apply a pre-calculated correction
formula to end up with the correct results. But this is how I imagine it...)

I guess it all depends on how accurate a profile is. If it's accurate, it should yield the correct target values.

As I imagine it now, the initial gamma curve can be different from the
target because you have to play with the luminance anyway. (Of course,
it should be somewhere close.)
Now, here is the trick:
- You can correct the gamma with the VGA LUT and mess with the numbers
during the gamut emulation to get back to the target luminance.
- Or you can start from the native limunance, do the gamut correction
and target the desired luminance.
Do you get it?

Sure. Both ways should yield the same target values, so I think the choice depends on where you can do the corrections with the least side-effects (which may be what you were pointing at in the first place, so, sorry if we are already on the same page :))

I can't really see it because I never examined any gamut conversion math
This is why I ask you...

Can't help with any math (probably for the better, because I suck at it ;)), but for LUT profiles, it can't be done using formulas alone. You have to use the lookup tables and also interpolate. For matrix profiles, it's obviously mostly matrix calculation.

 >>> I think I completely missed the fact that you want to watch « movies ».
Forget everything I just wrote L

Why? Color management is always the same in theory. The targets and the
process may differ but the idea is the same: achieve your target (mostly
one of the standardized) characteristic(s)...

With the exception that imho the whitepoint for movie viewing doesn't need to be something specific as long as you can perceive it as 'white' (and don't need to match several displays), while for graphic work you rather sooner than later arrive at a point where you want (or need to) communicate color, match a display to a known good print/proof/other display, etc.

Florian Höch

Other related posts: