[argyllcms] Re: bin/average: averaging and possible outlier elimination for three or more .ti3 sets?

  • From: Craig Ringer <craig@xxxxxxxxxxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 31 Aug 2009 09:58:31 +0800

On Sat, 2009-08-29 at 22:56 +0100, Alastair M. Robinson wrote:

> Craig Ringer wrote:
> 
> > 184: 37.428620 -5.000581 0.024627 <=> 78.810882 -2.002915 60.115930  de 
> > 73.023573
> > 732: 79.038963 -1.977274 60.090194 <=> 36.299104 -1.068111 -7.373731  de 
> > 79.868037
> 
> > See how the L* in particular looks almost neatly transposed between the 
> > values?
> 
> Yeah, that's odd - how close to each other physically were these two 
> patches?

184 and 732 are CD6 and CD5 respectively, so it's clearly transposition.
This seems rather odd in a strip read, though - how would two
sequentially read patches in a strip get transposed like this?

By the way, as I now have five .ti3 sample sets, I've taken the three
with lowest dE between them and run `average' on them:

   average sampleA sampleB AB.ti3
   average AB.ti3 sampleC.ti3 ABC.ti3
   rm AB.ti3

then built a profile:

  colprof -v -y -A "Rural Press" \
  -M "45gsm Nornews No FWA, PANPA Spec" \
  -D "Rural Press 45gsm Nornews" -C "Craig Ringer" \
  -qm -l 240 -L 80 ruralpress1

but the result was *awful*. Spectacularly, totally, utterly broken. If I
use it in Photoshop's "Proof Setup", for example, the displayed image is
mostly flat grey. If I turn on paper colour simulation it's mostly dark
black-ish grey. (I can provide the .ti1/.ti2/.ti3 files and .icc if
anyone's interested, but don't expect anyone to waste their time
unnecessarily).

Could this be caused by averaging more than two .ti3 files sequentially
like this? I haven't gone digging into the data yet to see, nor have I
tried building a profile from just the `AB' average.

A profile generated from just one of the sample sets seems to be a lot
saner. It'll be hard to say how good it is until I send output from it
to the press, but for previewing jobs that've already been printed it
seems to be a decent match. The on-screen press simulation (black ink
and paper colour simulation in Acrobat Pro) seems fairly sensible as
well.

I'd like to thank Graham and the others who've worked on Argyll for
making this software available as open source. It's been a huge help to
me as I learn and it's great having such flexible, powerful tools.

--
Craig Ringer


Other related posts: