[argyllcms] Re: ArgyllCMS: shadow details and blue color issues. What I'm doing incorrectly?

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 22 Oct 2012 18:33:19 +1100

Xavi wrote:
> Dear Graeme.
> I enjoy reading your replies although I discover that my knowledge of color
> management is very basic.
> All the images that I use to test the printer profiles, are mainly labelled
> with the AdobeRGB and the others with the sRGB color spaces. Then, seems
> that this is not the reason of color change.
> As I commented to you,  I have changed all the cartridges of my Pro9000 to
> use another brand of ink, to have more chances to discover what is
> happening. Then I can't use at this moment, the printer with the ink of the
> example of ICM that you used in your icclu example
> (Argyllv140_396parxes48Gris3Blanc.icm).
> I must to use the MP540 printer with the
> "MP_B01_A4_HPEverydayBrilloIIQ1Dif_NoVPre_396p64g3bArgyllv140_121011Opcio_qm
> _ACPU.icm". Sorry to the extent name, but I have created a lot of profiles
> and the name was growing trying to identify them.
> I have do the same example, than you show me, using this ICM:
> .\icclu -fb -ip
> MP_B01_A4_HPEverydayBrilloIIQ1Dif_NoVPre_396p64g3bArgyllv140_121011Opcio_qm_
> ACPU.icm
> 50
> 50.000000 0.000000 0.000000 [Lab] -> Lut -> 0.563818 0.500477 0.630465 [RGB]
> .\icclu -ff -ir
> MP_B01_A4_HPEverydayBrilloIIQ1Dif_NoVPre_396p64g3bArgyllv140_121011Opcio_qm_
> ACPU.icm
> 0.563818 0.500477 0.630465
> 0.563818 0.500477 0.630465 [RGB] -> Lut -> 52.109114 0.326121 -0.977612
> [Lab]


> I don't know if the reading with an spectro some patches, is a good way to
> show you what I obtain on prints but I have print a simulation of the
> patches of a colorchecker card and I have read the results with a spectro.

Without knowing the workflow of how those colorchecker values
made their way to the print (ie., measured colorchecker values
reproduced using absolute colorimetric reproduction ?),
I'm not sure this leads anywhere.

It's far better to directly check critical spot colors against the
profile. For instance, pick a neutral color (say Lab 50,0,0),
then look it up using the absolute colorimetric B2A table,
then print that RGB value using the same print path as
the test chart. Measure it. Check the delta E against the target.

You can also then check the profiles self consistency. How much
different is the relative colorimetric B2A for a given neutral
color compared to xicclu -fif ?

[Note - an instrument measurement by default will correspond
 to absolute colorimetric rendering intent. ]

If the colorimetric model of the profile matches the device
reasonably well, then you can look into the behaviour of
the perceptual rendering (as I was doing using the icclu).

> Applyiing the icclu example that you show me to the values of blue color
> patch:
> .\icclu -fb -ip
> MP_B01_A4_HPEverydayBrilloIIQ1Dif_NoVPre_396p64g3bArgyllv140_121011Opcio_qm_
> ACPU.icm
> 28.87   14.81   -50.15
> 28.870000 14.810000 -50.150000 [Lab] -> Lut -> 0.358232 0.399245 0.879326
> [RGB]
> .\icclu -ff -ir
> MP_B01_A4_HPEverydayBrilloIIQ1Dif_NoVPre_396p64g3bArgyllv140_121011Opcio_qm_
> ACPU.icm
> 0.358232 0.399245 0.879326
> 0.358232 0.399245 0.879326 [RGB] -> Lut -> 39.085966 11.557139 -49.964686
> [Lab]
> I must obtain, on paper and with perceptual intent, a patch with 39.09 11.56
> -49.96 Lab values, but I read a patch with 39,35      4,45    -48,45 Lab
> values. That's a deltaE76 of 11,77.

It's not that reasonable to be looking at delta E's of non-neutral
colors for perceptual rendering, unless you really understand what's going on.
By definition it is not colorimetrically accurate - it aims at a pleasing
reproduction that tries to preserve the look of the original (ie. not
clipping) while staying within gamut.

Graeme Gill.

Other related posts: