[Lcms-user][OpenICC] vgct tags

[Kai-Uwe Behrmann]

Marti and Graeme,

thanks for sharing your thoughts.

I cc to OpenICC, as the vcgt in littleCMS reading may be of interesst 
there too. 
More below I write about the Genereal aspect of this particular issue. 
The original thread is partitial included at bottom.

a practical view:
yesterday I showed some jpegs from a digital camera to friends. I opened 
gqview, a popular image browser under linux, and could see the images 
there in wrong colours. I draged them all to CinePaint to view them colour 
managed. My X setup loads automatically the vcgt in a monitor profile from 
a (minimalistic) hardware database.

Currently, as in CinePaint is no reading of the monitor profile in X atom,
I must enshure myself the profiles match. Even with setting the profile 
from a central configuration this is not shure, as it may get changed 
during runtime. This incompletenes is probably the point you make here.

On the other side the decoration looks with vcgt much more neutral. I have 
allmost all decorations set to some shade of gray. The appearence is more 
balanced. 
Now my point, and possibly the one of Hal, here is, currently 
the majority of desktops suffer from uncalibrated 8-bit LUT capable 
graficcard/monitor combinations. The vcgt enshures a minimum of 
calibration. It is mostly like an linearised printer. No one want missing 
it, even being not in a final (profiled) state.

The same applies for printer and other devices. They lack in allmost any 
case driver specific information, which is worse as it make the profiles 
in an high degree application dependent and opaque to most other
applications. This is worst case to an open minded environment. 


Generally:
Vcgt is one step to add more of the hardware configuration information 
to one place - the profile. My doubt is the current ICC approach prohibits 
inclusion of more hardware information to make the configuration complete. 
This is due to the non changing policy of many profilers expressed in 
copyrights.

As I understand, Microsoft will differ to osX in the way to set up 
things, as well as Linux will. The ICC did not try to cover such 
differences under one cover. If we find a way to include Linux OS 
specifics into the profile I would like to support it. 
Option A
is we have to create an meta format covering the hardware configuration + 
the unchangeable profile. As I see it all OSes goes currently that way.

Option B
is device profiles could have an general configuration tag 
containing:
o driver identifier (which every application can eighter understand or 
  reject the whole profile)
o a set of options specific to that driver

This enshures an application knows if it is capable to handle a certain 
profile. And the application knows how to talk to the driver with a
specified configuration set (including profile) or can look up a 
suitable profile by given options.

I know this is something technical. And most users want to stay with a 
configuration and dont change it once it works. The difficulty I see here 
is, this sheme is too static.

Further I have to look at the outputResponseTag Graeme mentions below.

regards
Kai-Uwe Behrmann
                                + development for color management 
                                + imaging / panoramas
                                + email: ku.b at gmx.de
                                + http://www.behrmann.name


Am 15.10.05, 10:11 +1000 schrieb Graeme Gill:

> Marti wrote:
> 
> > lcms does currently NOT implement it because it is not on ICC spec.
> > Alas, I'm a bit critical on such practices. Using this kind of stuff will
> > make profiles "active" in the sense they are no longer a characterization
> > of the hardware but a way to set specific configurations.
> 
> You're quite right about this. vgct is not the correct approach
> to calibration. What should happen is that the profile should
> contain the desired device response in the outputResponseTag,
> and some separate system should calibrate the hardware to
> make sure it meets the expected response for a given display
> profile. This provides a mechanism to track device drift.
> 
> Unfortunately we're stuck with software systems that expect
> the vgct tag to be present (Apple systems won't load the
> profile if the tag is missing), even though it no longer
> captures all the settings of a modern display system, such
> as the monitor controls etc., and is the wrong information
> to store in a profile ....
> 
> Graeme Gill.
>