[argyllcms] Re: Monitor profiles and their (black point?) "readability"

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 19 Jun 2008 10:10:23 +1000

Klaus Karcher wrote:
As the Argyll profile has a media black point tag with Y > 0, everything below the black point will get clipped in colorimetric transforms without BPC. The GMB profile has no black point tag, therefore the CMM assumes XYZ 0/0/0 as black point. When yo convert RGB 0/0/0 from the Argyll to the GMB profile without BPC, you will get RGB 17/17/17, when you convert from GMB to Argyll, source values below 17/17/17 will get clipped.

Hmm. The black point tag in Argyll profiles simply reflects the behavior
of the device. The black point is still the same even if the tag
is missing. Many artificial RGB profiles have a black point of 0
(ie. sRGB etc). To smoothly convert between such spaces and a real
world profile, some form of gamut mapping is really required.

Black Point Compensation is a crude form of run time gamut mapping.

Another approach is to setup the Argyll profile with appropriate
gamut mapping in the B2A tables for a black point of 0 source.
This assumes that the programs in question will/can use perceptual
or saturation intent.

If they are using relative colorimetric intent, then the
black clipping behavior is exactly what they are requesting,
and if this behavior is objectionable, then I think they probably
have a bug. Perhaps they should use BPC.

I'm not sure why other profiles may not have this behavior, unless
the profile creation is somehow distorting the device characterization
to make it appear as if the black point is zero. Since the ICC
profile is meant to reflect the measured behavior of the device,
I can't see that this is a valid approach.
[To investigate this further would require specific profile examples
 and scenarios.]

Kodak proposed to delete the mediaBlackPointTag and AFAIK this proposal is already approved by the ICC. Whether this decision makes sense is a different kettle of fish.

No sense at all is the answer. The black point and white point
are pretty basic media information, and essential when doing
gamut mapping. A black point can be created from the A2B colorimetric
table, but it may be different to what the profile creator (who
is likely to take the most care about things and has more information
available) would create, and is also difficult to determine for CMYK
without ink limit information, which the ICC format lacks.

Graeme Gill.

Other related posts: