Re: ImagePrint
Re: ImagePrint
- Subject: Re: ImagePrint
- From: Graeme Gill <email@hidden>
- Date: Fri, 24 Jan 2003 16:00:37 +1100
Ernst Dinkla wrote:
>
Could someone comment on the following observations:
>
>
When I load an ImagePrint "RGB" printer profile in ICCinspect
>
(www.littlecms.com) I see lots of slopes, shoulders, hiccups etc in the in-
>
and output curves. None of that is seen in a custom "RGB" profile made with
>
Profiler Pro for the same ink and media setting in ImagePrint. The two
>
profiles gamut shapes look almost identical in the 3D profile viewer of the
>
Wasatch SoftRip. I got the two profiles from James Irelan because he had
>
difficulty to get blacks without lighter streaks in them with the custom
>
profile. No ImagePrint here but I'm curious what is so unique about that
>
driver and its profiles.
>
>
Checking other "RGB" printer profiles from various sources don't show any of
>
those typical curves in ICCinspect or just a small contrast shift in one of
>
the curves.
>
>
I'm on thin ice here but it looks like an attempt to give some inklimit, etc
>
instructions normally found in a CMYK printerprofiles by bending the curves
>
in the "RGB" profile. Most likely it also has linearisation compensation as
>
the hiccups are probably at the point where the Cc, Mm, Kk inks switch.
Creating gamut views of profiles is a difficult business. There
are all sorts of trade-offs one makes between speed, accuracy and
correctness.
One approach assumes that the gamut surface is always coincident
with the extreme device values, and that the surface has the same
topology in input and output space. If this is true, then
it's easy to get accurate, continuous looking gamut surfaces
by mapping a description of the device value surface into PCS.
As soon as you move to CMYK, the assumption about surface continuity
can no longer be true (since in device space we're talking about
the surface of a 4 dimensional object, while the gamut is the
surface of the 3 dimensional PCS values.)
One approach to address both assumptions is to sample the
profile (or in the case of ICC clut based profiles, use the
clut node values as sample points), and then in some
fashion, decide which points form the gamut surface,
and which lie below it. This is not easy, and some
sort of heuristic tuning parameter is needed as a decision
criteria. You then need to form of geometric surface description
out of the points you've decided make up the gamut surface.
This approach is more robust and universal, but is subject
to displaying various artefacts due to the judgement criteria,
surface description creation etc.
The issues I've described might go some way towards explaining why
different gamut viewing applications give different looking results -
the artefacts they produce have different sensitivities to
the details of different profiles. Viewed with my Argyll toolset,
the Colorbyte gamuts look about "normal" for an RGB printing profile.
I've observed before on this list that the Colorbyte profiles do
seem a little different and "non-standard" in some ways. The gamut
the profile indicates gave PCS values well outside the actual values
the devices in question are able to achieve (at least on the
Epson 10000). The black generation also seems to be extremely
primitive (but I'm not clear on whether the Colorbyte RIP is
doing the black generation, or whether they are driving the
Epson in RGB mode). The "normal looking shape with a cone
sticking out of it at the black end" is a symptom of using
a simplistic black generation algorithm of the type
K = min(C, M, Y)
C = C - K
M = M - K
Y = Y - K
This sacrifices a great deal of the devices possible gamut
in the saturated dark color region.
Graeme Gill.
_______________________________________________
colorsync-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/colorsync-users
Do not post admin requests to the list. They will be ignored.