Re: Pre-Linearization or Iterative Profiling (WAS Re: Printer profile mis-shape?)
Re: Pre-Linearization or Iterative Profiling (WAS Re: Printer profile mis-shape?)
- Subject: Re: Pre-Linearization or Iterative Profiling (WAS Re: Printer profile mis-shape?)
- From: Graeme Gill <email@hidden>
- Date: Wed, 30 Aug 2006 00:38:05 +1000
Roger Breton wrote:
Graeme, you seem to follow a similar approach in argyll in that you start
out with an XYZ target definition?
The equivalent feature in Argyll is aimed at improving the
multi-dimensional sampling efficiency. It is not aimed at,
and would not be as effective as a per-channel curves in
addressing things like inkjet dot gain, as I would address
this in a calibration system, rather than the profiling system.
To be able to do an optimized test chart, it is necessary to
bootstrap into it. To accomplish that, an existing similar,
or previous profile is needed, or a profile has to be made
using (say) a non-optimised/non-adaptive (probably smaller)
test chart.
Even for non-adaptive test chart, it's possible
to sample the device color space a lot more efficiently than
a rectangular grid - there are several geometric arrangements
that fill space more efficiently than that, for instance
the body centred cubic grid, or itteratively optimized
space filling algorithm.
For an adaptive chart, an optimal sampling algorithm will
do something like try and pick sample points in an
arrangement that minimizes the largest error anywhere
in the gamut, between the CIE value interpolated by the
profile created by the sample points, and the actual
CIE value the device produces at that point.
The algorithm I'm using at the moment for an adaptive
chart is to choose a series of sampling points in an
order that locates the next furthest point in device space
from any existing sample point, that has the largest
error between the interpolated CIE value for that point,
and the CIE value at that point predicted by the bootstrap
profile. It's not as refined as I would like, but it
heads in the right direction.
Graeme Gill.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden