[argyllcms] Re: Am I stupid or is it something else... my color management sucks.

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 15 Apr 2010 14:54:17 +1000

Michel Joly de Lotbinière wrote:
I use a DTP94  colourimeter, and the only way to achieve neutral dark
grey tones on my notebook's decent but average display is to make a
Single Curve + Matrix profile: the RGB Curves + Matrix always have a
slight reddish cast in the RGB(0,0,0) to RGB(40,40,40) range of grey
tones, whether I use the Windows or LCMS colour management engines.

Most likely because of the different colors of white and black on
an LCD. By default dispcal attempts to get a deep black point
on the LCD by compromizing the neutrality (compared to white).

If using dispcal -k1 cures the neutrality problem (at the cost of
a not so dark black point), then this is the explanation.

One way of improving things may be to pick a white chromaticity
that is closer to the chromaticity of the black.

Another might be to fiddle the -A parameter. A bigger number may
give a deep black with less compromize in the visible neutrality.
In my experience it can create worse viewing angle dependent effects though.
(LCD's are far from the perfect display device).

If I could figure out how all the black level offset, correction rate
and percentage adjustments actually work (e.g. does a higher numerical
correction rate imply that dark grey tones are pushed faster towards
the display's *native* colour temperature or towards the assigned
*target* colour temperature? Is anything done via the correction rate
if 0% correction percentage is specified, as is the default for
LCDs?), then I might be able to achieve better accuracy with a
3xCurve+Matrix profile.

The blend rate doesn't do anything if there is 100% correction,
since then there is nothing to blend between. It only has an
effect if the black color is different from the white color
because no attempt is being made to make them the same.

The dispcal help text is really confusing here in terms of answering
the question: if the dark greys of a step wedge observed in a
colour-manged application using the final profile (perceptual intent)

The profile adds another level of color manipulation, and so
the dispcal help text can't help you there, it is only
relevant to the calibration.

The profile reflects the behaviour of the device,
and the CMM is free to use that as it sees fit. It may decide to
use the black point as given, or it may not.

produced by colprof are biased on the red/cyan axis when viewed on an
LCD display, but they are neutral in the calibration produced by
dispcal (as judged by observing a neutral step wedge in a non-colour
managed application), what adjustments can I make in dispcal to
attempt to correct that?

None directly, since it's a profile issue. You need to look to the
profile and CMM first and understand what they are doing, before
deciding whether changing the calibration is a path than may help.

Maybe DTP94 (and Spyder3?) readings of dark colours are simply less
accurate than bright colours, and this skews the profile calculations,
even if the calibration LUT curve produces neutral dark greys.

That's not typically the case. Often readings get noisy.

Graeme Gill.

Other related posts: