[argyllcms] Re: Profile generation with perceptual intent with compensation for paper white
- From: Henrik Olsen <henrikolsen@xxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 22 Sep 2017 10:10:07 +0200
On 22 Sep 2017, at 02.47, Graeme Gill <graeme@xxxxxxxxxxxxx> wrote:
Henrik Olsen wrote:
But I guess we also adapt to the surrounding light and known colors around
us, and
therefore can spot a bluish paper/print lacking the relatively warmer tones
I edited
the images in.
Not usually, no. Adaptation is complex, part of it being in how the receptors
in
the retina automatically adapt, part being how the opponent signals derived
from the receptors make us sensitive to white balance, and ultimately there
is a cognitive component striving for Color Constancy. In my experience
this all adds up to an ability to almost completely disregard the
color of the paper in the context of comprehending an image, even with
highly tinted papers. You can manipulate these mechanisms though, by
introducing other white cues, or eliminating the perception of there being
an image.
So the question I would be asking is: if your viewing environment
is making you aware of the slightly blue nature of the paper, does it
reflect a realistic viewing condition ? I would suspect not.
If evaluating the print right where it’s supposed to hang, it can’t be more
realistic than that. I don’t prefer the bluish rendition of colors edited to a
different tone. It doesn’t appear as intended. But does if printed on neutral
paper.
I don’t see my prints only in a viewing booth, or with a mask blocking
everything else out around me. If so, I’m sure human white point adaption is
likely doing it’s thing. I evaluate my prints on the table, in my hands, on the
walls - with natural surroundings not usually blue tinted. My walls are close
to neutral, and my preferred passepartout is too.
[ And I am very familiar with FWA loaded papers, having dealt with them
extensively in the context of color copier color management. 99.99% of
the office copy paper that you will come across is low quality paper
loaded up with FWA. As I write this, my desk is littered with such
paper (M0 D50 b* = -9.25) , and none of it looks blue. On the contrary,
if I introduce a piece of Tyvek which I know to be a very highly reflective,
uniformly spectrally reflective white into my field of view,
it looks quite yellow. Now if I filled my desk with pieces of
Tyvek and introduced one piece of Office paper, I'm sure the
reverse would be true :-) ]
Yep. And I’m mostly surrounded with neutral colors on the walls. Wouldn’t feel
nice with a b* of -9.25 on the walls :). So when walls and passepartout are
neutral, and I’m not surrounded by loads of bluish copy paper, I really do
observe the images being blue tinted (which they are) when printing rel. col.
on typical blue RC paper. I don’t have a white balance spot meter in my brain
making the print look as intended because I white point adapt to a bluish paper
border - doesn’t work for me. The minute the print is replaced with a neutral
toned one (replicating the colors I edited as), it clicks for me. So I’d still
like the option of having my colors come out more as defined, with the
warmth/tint intended, not looking bluish/dull/lifeless. Likely I really only
achive this perfectly on neutral paper, but I’d like to emulate that on bluish
paper as well.
Also, the issue isn’t just related to FWA/OBA. The print looks bluish and dull
to me, even in indoor lighting with not much OBA triggering by UV, and the
papers often have an unwanted dip in reflactance around 500-650nm, besides the
OBA hump around 440nm. Getting it triggered by UV makes it even worse, like in
daylight. Getting a more 1:1 rendering compared to edits, enjoys me. Easy on a
neutral paper like HM Silk Baryta with nice flat reflectance. Getting closer to
that rendering on bluish stock is now an option I’d like to investigate, and
Hahnemühle seems to be getting that point, and ProfileMaker and others, making
perceptual profiles available that will attempt to neutralize the rendering.
So I see great value in an ability
to control the neutrality of perceptual intent, but cannot find a way to do
so with
ArgyllCMS - which I otherwise love.
I've suggested a means of achieving what you want (collink -iaw) - give it a
go, and see
if it works the way you expect.
I’ll check it out. Went there before, but will try it again. But ‘aw' intent
will clip my shadows brutally. Perhaps ‘la’ intent could work better. As
ProfileMaker describes it, I’d like: “...relative colorimetry in highlight
areas and absolute colorimetry with shadow compensation in the remaining
colors.”
I notice ArgyllCMS documentation says "The pa intent, Perceptual Appearance
uses "knee" type 3 Dimensional compression to make the source gamut fit within
the destination gamut. As much as possible, clipping is avoided, hues and the
overall appearance is maintained. The white point is not mapped from source to
destination, allowing the apperance parameters to alter the chromatic mapping.”
The last sentence gets my interest. What are those "apperance parameters”, and
how to alter them. Is the "Grey axis alignment factor” (reported during colprof
for percep) part of those parameters? And what does this factor do? Could it
used to get what I want, even if requiring a recompile in case it’s not end
user tweakable?
/Henrik
Other related posts: