[argyllcms] Re: Explain what happened
- From: Graeme Gill <graeme@xxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Wed, 03 Dec 2008 12:52:14 +1100
Leonard Evens wrote:
I used
dispcal -v -yl -0 TargetA
with an Eye-One LT
I ignored the first set of numbered instructions because there are no
controls that I can adjust---which may have been a mistake---, and I
went directly to the last two steps: the check of ambient light and
making the profile. It did produce a profile TargetA.icc.
I'm not sure why that happened. It shouldn't have produced a profile,
since you didn't give it a -o parameter. (-O sets a profile description).
Are you sure that your command line above is accurate, and that
it wasn't in fact:
dispcal -v -yl -o TargetA
???
Note that using dispcal -o, the monitor is both calibrated
and then the calibrated monitor is profiled (ie. characterized).
Naturally, the profile is only valid for the monitor in its
calibrated state, that's why the calibration curves are (by
convention) stored in the profile, so that the monitor
can be set to the calibrated state whenever the profile
is to be used.
I assumed on the basis of what I thought I understood that the
calibration part, i.e., loading the video LUT was irrelevant and that
the profile would be made under the assumption that it would do all the
work.
I'm not sure why you thought that. The whole point of device characterization
(profiling) is to record and model the color behaviour of the device.
This means creating a equivalence between device color values and
device independent (CIE - ie. human perception) color values. That mapping is
then represented by an ICC profile model. Calibration changes the behaviour of
a device by altering the way that device color values are
converted to/from actual (human perception) color. So obviously a calibration
state affects the profile. For a profile to be valid, it must
be in the same calibration state it was when it was characterized.
You used dispcal to both calibrate and profile, so the profile
is only valid if the monitor is put in its calibrated state.
I then told gimp that TargetA.icc was the monitor profile. But
that had no effect whatsoever on the colors that gimp put on the screen
in an image window.
Sorry, I don't know how the Gimp handles (or doesn't handle) color profiles.
If nothing changed, then either the profile was the same as the one
the Gimp was already using, or there is no color space definition (profile)
set for the image being displayed, or the Gimp isn't working the way one would
expect when informing it of the monitor profile.
[One wouldn't normally expect an application to set the monitor
to its calibrated state - this is more of an overall system
concern, not something an application that uses profiles should do,
hence the use of calibration loaders such as dispwin or xcalib
on operating systems that don't have this built in.]
So I tried
xcalib TargetA.icc
and that immediately turned the background (which I had previously set
to R 128, G 128, B 128) to neutral gray. (Before it has been the bluish
gray you sometimes get on un calibrated laptops.) In other words a LUT
was in fact loaded in the video card.
Yes, xcalib loaded the calibration curves stored in the ICC profile
onto the monitor, (hopefully) putting it in the calibrated state
that dispcal created and then used for characterization.
It doesn't explain why the Gimp didn't change its display
when you informed it of the monitor profile though. If the
Gimp is actually doing color management, then it will have a
definition of the color space of the pixels to be displayed,
and will use that together with the monitor profile to transform
the pixel values for display (ie., by linking the two profile
together and transforming from the image color space to the
calibrated monitor space).
So I clearly don't really understand the distinction between calibrating
and profiling.
Hopefully the above discussion helps.
And I don't really understand what gimp color
management is doing.
I'm not sure I can help you with that. Perhaps the Gimp
mailing list might be a better forum ?
Graeme Gill.
Other related posts:
