[argyllcms] Re: Camera matrix profile, adding ti3 perfect white data set

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 01 Feb 2010 14:33:28 +1100

Elle Stone wrote:

Could someone please explain what "media white point" means when the
thing being profiled is a digital camera?


ICC profile internal data is always relative colorimetric, which
means that the media white always appears to be D50 so that the
white point is common in the PCS. So the white point tag contains the
absolute colorimetric measurement of the white point, so
that the relative colorimetric profile data can be converted
back to absolute colorimetric for the absolute colorimetric intent.

For an input profile the test chart is assumed to represent
the media, therefore the white and black points are
set to the white and black of the test chart. This way
the relative colorimetric interpretation of that media (the
test chart) on that device has a D50 white.

Now in reality the test chart may not be of the same media as
that actually used with that input device, creating practical
problems with white and black points.

The -u flag is my attempt to provide a workaround for these
issues, where the white and black points will be determined
and adjusted after color management, rather than part of color management.


I did notice in 1.1.0 that the media white point for a matrix profile
is set to D50, rather than being based on readings from the brightest
patch - normally GS00 in an it8 target, as was the case in V1.0.4 (I
think the brightest patch is what determined the media white point -
yes?). Are you saying this a "bug"? I thought it was a feature. Or are
you talking about something else altogether?


Yes, that's a bug.


For a digital camera, is "media white point" the color of the light
hitting the target? In which case perhaps the media white point
shouldn't be based on GS00 but rather on the most neutral patch on the
target, which for my particular it8 chart is GS06, assuming my
batch-averaged chart doesn't have a gross error for GS06 which it
doesn't seem to.  In fact I'd take a guess that the neutral patch with
the highest error on my chart is actually GS00 - the L* part always
comes out way off in profcheck and the patch itself is far from being
really neutral.


I'm not sure that a media white point makes much sense for a camera, unless
it's part of a camera stand for photographing documents.


Also, the black point for a matrix profile in V1.1.0 is zero (with or
without an artificial black point added to the data set), rather than
being based on the darkest patch on the it8 target as it was in
V1.0.4. I had already decided to get rid of the non-zero black point
in my camera profile, simply to prevent software from doing unwanted
black point compensation (I'd rather do my own). So is this new zero
black point in V1.1.0 a feature?


No, it's the same bug.

(note that -u will set a D50 white point and zero black point, since it's
 a "non-ICC standard absolute colorimetric intent all the time" option.)

Graeme Gill.

Other related posts:

  1. Home
  2. argyllcms
  3. Archive
  4. 01-2010