[argyllcms] Re: Can anyone help with a weird OSX problem?

  • From: Florian Höch <lists+argyllcms@xxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Sat, 26 Jul 2014 00:39:03 +0200

Am 25.07.2014 um 22:19 schrieb Quartz:
> Here are two images comparing it to sRGB and AdobeRGB. My display's plot 
> is the white skeleton outline, and you can see how blue (and to a lesser 
> extent yellow) are sticking out.
> http://www.sneakertech.com/-/srgb.png
> http://www.sneakertech.com/-/adobergb.png

A bit hard to see on the side view how much it pokes out.

>> Out of curiosity,
>> can you send me the display profile? (or alternatively upload it
>> somewhere and post a link)
> http://www.sneakertech.com/-/display.zip

I'm using Argyll's xicclu to examine the conversion of sRGB blue to your
display profile (using relative colorimetric intent).

First I get the sRGB blue L*a*b* values:

$ xicclu -ir -pl -s255 sRGB.icm
0 0 255 [RGB] -> MatrixFwd -> 29.565320 68.291889 -112.034964 [Lab]

Now the closest match using your display profile:

$ xicclu -fb -ir -pl -s255 display1-f70bfba8402f41789d351e95fa3e2345.icc
29.565320 68.291889 -112.034964 [Lab] -> MatrixBwd -> 96.401697 0.000000
255.000000 [RGB] (clip)

That is almost exactly the RGB values you measured in Photoshop
(actually the difference is probably due to integer rounding). The
"(clip)" means the color is out of gamut of your display (expected).

And now back from above display RGB to L*a*b*:

$ xicclu -ir -pl -s255 display1-f70bfba8402f41789d351e95fa3e2345.icc
96.401697 0.000000 255.000000 [RGB] -> MatrixFwd -> 43.817886 29.633722
-83.880413 [Lab]

That L*a*b* value has a delta E 1976 of 49.9 vs. the original sRGB blue!
Not great (delta E 2000 is 16.8, better, but still not great).
But that color isn't purple.

> This is one of the last profiles I created, using dispcalGUI+argyle (and 
> the profile the above pictures are based on). Since I've been messing 
> with options trying to figure things out, I don't 100% remember what I 
> set to create this, other than I'm pretty sure it's a 3-curve matrix 
> with native white point.

You can check this easily by selecting the profile under "Settings"
(it's a 3x curves + matrix profile), or if you deleted it from there,
dragging & dropping the file on the dispcalGUI window.

> From what I 
> read online, Photoshop under windows on the same hardware may not be 
> affected either, but I haven't had time to test that personally.

Technically that seems unlikely (using the same profile).

> What I was getting at is that I think it's a software issue rather than 
> a busted screen/colorimeter.

I think it's neither a software issue, nor are your screen or
colorimeter busted. You basically have two issues:

- Very limited gamut of the MacBook screen (nothing you can do about
that as it's a hardware limitation)
- i1 Display 2 which has aged and is potentially not very accurate, due
to the reasons mentioned (this could be worked-around with a colorimeter
correction, and/or by calibrating to the native white point)

> Just to make sure I understand you correctly; a correction matrix would 
> be specific to the colorimeter I'm using (so I couldn't just download 
> one online), and in order to create one I'd need to get ahold of more 
> hardware?

Basically, yes. You could try finding one online, there's just no
guarantee it'll work well with your specific instrument/display combo,
but may be better than nothing.

Another option that I forgot to mention is that you can import
corrections from other display profiling software in dispcalGUI (look in
the "Tools" menu). Quato's iColor Display has a generic correction for
i1 Display 2 and Apple's Cinema Display with white LED backlight that
can be imported.
The Cinema Display has a gamut that is much closer to sRGB than the
MacBook screen, so I'm not sure if it'll help, but it may be worth a try.

I'd try these things (in order, and see if any of those improve things):

- Calibrate to native white point (& profile)
- Create a XYZ LUT + swapped matrix profile (with default settings)
instead of curves & matrix
- Use the generic colorimeter correction mentioned above

I'd not recommend buying a spectro just for the sake of creating a
correction for your i1 Display2. It may be an option if you plan to
profile other devices (e.g. printers). If you want to replace the i1D2
with a better colorimeter that has shielded filters and comes with a
generic correction for white LED in its software (which can also be
imported), the ColorMunki Display (or the more expensive, but also a lot
faster i1 Display Pro) may be an option.

Florian Höch

Other related posts: