[argyllcms] Re: Average bug/Density curves/ xicclu -g predictability issue

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Wed, 19 Jan 2011 09:28:00 +1100

Elena [service address] wrote:

Mmm... really, why didn't I think of that before ?!
Unless having duplicate SAMPLE_IDs in the ti3 might lead to some
unpredictable side effect...

The Sample_ID is ignored by colprof.

Yes, but since I see that colprof generates shaper curves that aren't
quite linear (see attached image) when channels are targeted at linear
density, I was wondering if my idea could improve things furtherly.
No more than one of my academical fixations :)

I wouldn't be to concerned. The curves are relatively smooth, and don't deviate
very much at the 50% mark. The important thing is that the resulting
per channel output is visually progressive, and roughly even.
(There is only so much that per channel curves can do anyway.)

always playing with xicclu K curves, I am a bit disappointed that a
predicted plot doesn't completely reflect what colprof will actually generate.

They are not exact, no, but typically they do at least resemble each other. You
need to keep in mind that the B2A table is an sampled version of the
fine grained plot you get out  xicclu -g, with some tweaks to make it
more like a least squares fit to the underlying inverse. The finite
resolution of the B2A table means that it tends to smooth the underlying
curve out. This may be good or bad. It's better of course if the underlying
curve is smooth to start with.

  xicclu -v -g -fif -ir -kp 0 0 1 .2 .5 -L100 -l300 test5.icm
  the plot: http://www.elenadomain.it/pub/argyll/predicted.png

  xicclu -v -g -fb -ir test5.icm
  the plot: http://www.elenadomain.it/pub/argyll/actual.png

They seem an excellent match - very pleasing.

Graeme Gill.

Other related posts: