[argyllcms] Re: What to include in bug reports?

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Wed, 14 Oct 2009 01:47:29 +1100

Daniel Bachler wrote:

The crash during profile generation with version 1.1 happens when I
start it from the command line as well

Hard to say if it's a known issue or not.

The other problem I noticed (several patches displaying repeatedly
while the patch counter goes up, then the patch counter staying the
same and patches changing colour during calibration) is hard to
reproduce. I am not sure if it happened this time again. The
calibration file does not show any two lines with identical values, so
maybe everything works as intended and I was seeing things...

I haven't seen this behavior on MSWindows :- it's possibly
something to do with the GUI.

is fine. I assume that this is because the prophoto colour space
extends outside the famous horseshoe of visible colours, correct?

Yes, the ProPhoto Blue is outside the spectrum locus, and CIECAM02
was formulated for real world colors, and doesn't handle those
near or beyond the spectrum locus. (I presented a paper at CIC
last year on some approaches to minimize this problem,
see <http://www.imaging.org/ScriptContent/store/epub.cfm?abstrid=42701>).

But
what I find strange is that it doesn't get mapped into the most
saturated blue range but instead to a visibly darker and (maybe) less
saturated area.

The problem is that it's not possible to know what color the blue is,
since it's not real. It's got almost zero luminance, so it must look
black, right ?

So what happens in practice is that it maps to whatever is closest in the
colorspace coordinates that it's represented in.

Is there a way around this issue? Is this an issue of
photoshop? Or is this transformation more complicated that it seems to
a novice in colour transformation like me?

You can probably work around this if you force Argyll to use
L*a*b* for it's mapping/clipping space, ie. by selecting the "rl"
gamut mapping intent.

Of course there shouldn't be a problem when mapping real world colors.

cheers,

Graeme Gill.

Other related posts: