[argyllcms] Re: Problem with calibration

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 02 Mar 2009 11:55:21 +1100

Fridtjof Busse wrote:
For example: If I use the profile produced by Argyll, many of the
numbers on the testchart [1] are invisible, while the profile from
basICColor shows even the "1" (though very weak).

There are many possible explanations for this, but without further
information it's hard to guess what your particular situation is.

One issue is that the test chart you refer to is an unknown.
It's unknown what situation it is intended for. What gamma is
it encoded in ? What display brightness is it intended for ?
What viewing environment is it intended for ? If any of
these things don't match your situation, the chart may be difficult to read.

It's also not clear whether you are talking about calibration
or profiled behaviour. The latter can be affected a great deal by
the way that profiles are linked (ie., whether the source to destination
black points are being mapped to each other.)

If you're talking about calibration only, then there are still
several variables that could explain the difference, such
as a difference in target parameters, even though you
say you've tried to match them. Note for instance that
Argyll defaults to a Gamma target of 2.4 on non-Apple machines,
a number suitable for viewing typically encoded images in a darkened
viewing environment. This is slightly higher than the defaults
of some other software, and results in a slightly higher contrast
that may make a chart intended for display at a lower gamma
hard to read in the shadows.

Also note the lack of agreement on how real world gamma's are
measured. The default 2.4 target in Argyll is an effective
gamma value, not a technical gamma value (ie. the curve
shape won't follow a power of 2.4 once the non-zero black
is accounted for, but the mid level brightness of the actual
curve used will match that of the theoretical ideal power curve.).

Any way to determine why the profiles are so different?

If it is the profile (rather than calibration), then the basic
measure of profile quality is how accurately it represents the
display behaviour. The usual way to evaluate this is to
measure the display (ie. generate a test chart and then
used dispread), and then verify against the profile

Graeme Gill.

Other related posts: