[argyllcms] Re: Targen options

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Tue, 04 Sep 2007 13:43:54 +1000

Ben Goren wrote:

Is Photoshop's  color engine  really that bad,  or is  Argyll that

I think that Photoshops CMM is actually very high quality, but the
difference is in using a device link, rather than linking the
A2B and B2A tables of two profile. Photoshop doesn't support device links.
Directly generating a device link from the A2B tables of the two
profiles just cuts out all sorts of information loss and inaccuracy,
as well as allowing a specific gamut mapping.

The device gamut fills almost 100% of the A2B table, since it is
indexed by device color value combinations. In contrast, the device
gamut is lucky if it occupies 20-30% of the B2A table, since it is
indexed by an L*a*b* cube, much of the volume of which are
"imaginary" colors that aren't realizable in the real world. This
means that for a given size, the B2A table is not able to hold
as much information, and is simply "coarser". Another issue
is that currently the Argyll B2A tables are probably not as
good as they could be - there are some tricks that can probably
improve their accuracy, and this is the project I am hoping to
complete to release with V0.70.

I've wished  for such on  a couple occasions. I don't  suppose you
have any suggestions for one that works with a Canon i9900 on OS X
(or a  free Unix (preferably  OpenBSD))? I sure haven't  found any
that cost less than four times  as much as the printer that looked
like they were any good....

That's the nature of such things. They are a more specialized product,
and someone has to pay for such things to be developed and supported
(I should know, since that's what I used to do !).

Something like the Linux inkjet drivers offers the scope to do this
(ie. Gutenprint), but it's not very well setup from the color point
of view in my opinion (too much of the old style hand tweaked curves
and equations, rather than modern instrument based calibration and
profiling), so it mightn't give any better control or result than
the built in RGB drivers.

I have a project to write a driver for the Epson R1800 I own, but
haven't found a chance to do much work on over the last 12 months. Such
a driver might be adapted to other printers, if the print protocol is

Graeme Gill.

Other related posts: