[argyllcms] Re: Perceptual intent

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Wed, 17 Dec 2014 11:20:51 +1100

robert@xxxxxxxxxxxxxxxxxx wrote:

> As per my email to you, I am using GamutVision from Imatest to view things
> like the 3D Lab gamut and black and white density response curve.  (I've
> resent the email as I can't show graphics on this mail list. The email also
> contains links to the profiles, Argyll .tix files etc).
> So my figure for the DMax is computed by GamutVision.

        I don't have GamutVision, and even if I did, I'm not sure
how useful it is to explore the ins and outs of how it does things,
rather than use tools that I know everything about.

> I really don't understand what you are showing me with your Fogra39 example.
> Here are results from xicclu for the Argyll and i1Profiler profiles for the
> same paper:
> H:\bin> xicclu -ff -ip iPF6400-Canson-Baryta-310-Argyll-2584.icc
> 0 0 0
> 0.000000 0.000000 0.000000 [RGB] -> Lut -> 3.060566 -0.115180 1.191879 [Lab]
> 1 1 1
> 1.000000 1.000000 1.000000 [RGB] -> Lut -> 100.000000 -0.000062 0.000060
> [Lab]
> H:\bin> xicclu -ff -ip iPF6400-Canson-Baryta-310-i1Profiler-2584.icm
> 0 0 0
> 0.000000 0.000000 0.000000 [RGB] -> Lut -> 0.000000 -0.093750 0.097656 [Lab]
> 1 1 1
> 1.000000 1.000000 1.000000 [RGB] -> Lut -> 100.000000 0.000000 0.000000
> [Lab]

You are exploring the A2B here, whereas my example was exploring the gamut
mapping from source space to destination by use of a device link.

I suspect that the i1Profiler perceptual table is conforming
to a different (undocumented) convention, one where the A2B output is stretched
to an unspecified intermediate gamut or luminance range. Argyll doesn't attempt
to do that - the perceptual & saturation A2B is always the same as the 
colorimetric A2B,
and the B2A maps from the source gamut specified in the profile creation,
to the device gamut.

See <http://www.argyllcms.com/doc/iccgamutmapping.html>

> An L* value of 0.0 from i1Profiler does seem a bit suspicious!  However the
> Argyll L* value of 3.0605 works out at a DMax of 2.47 according to Bruce
> Lindbloom's companding calculator, so it seems that GamutVision is getting
> things wrong (DMax of 2).  This is assuming xicclu is giving the right value
> and that I'm using the command correctly.

Argyll's model of this devices behavior is that it has an L* of 3
for black. The actual patch value may be slightly different to this -
the profile represents a "consensus" of the device behavior derived
from the nearby measurement points. It is unlikely that a real printer
has an L* of 0, and supports my guess that the i1Profiler perceptual A2B
is expanding to some (unspecified) intermediate gamut or luminance range.

> H:\bin> xicclu -ff -ir iPF6400-Canson-Baryta-310-i1Profiler-2584.icm
> 0 0 0
> 0.000000 0.000000 0.000000 [RGB] -> Lut -> 3.737745 0.015625 0.765625 [Lab]
> 1 1 1
> 1.000000 1.000000 1.000000 [RGB] -> Lut -> 100.000000 0.000000 0.000000
> [Lab]

> Are there any other tests that you can suggest to get to the bottom of this?

The number of different tables in a cLUT profile (6), combined with the
varied implementations of perceptual and saturation intents, leads
to lots of different possible gamuts. To compare things, you need
to know that you are comparing apples to apples. That's why I
look at the colorimetric gamut - it's the bedrock. Other tables
gamuts (B2A, or perceptual & saturation A2B) could have various
differences due to inversion limits or gamut mapping effects.

> Regarding BetaRGB:  I realize that it has a large gamut, but so does my
> printer. I normally print using Relative so it isn't such an issue; when I
> find that Perceptual gives a better result I use the smallest possible
> working space. When I make a new profile I always check it with GamutVision
> to make sure that there aren't any obvious problems and this is how I
> noticed that there appeared to be a problem with the Argyll perceptual
> profiles.

As long as you are conscious that Argyll's perceptual conversions are
compressing from your source gamut to the printer, and that therefore
to not loose colorfulness, your images need to occupy the source gamut.

Graeme Gill.

Other related posts: