Re: Real world experience w/ GMG and Oris RIPs
Re: Real world experience w/ GMG and Oris RIPs
- Subject: Re: Real world experience w/ GMG and Oris RIPs
- From: Roberto Michelena <email@hidden>
- Date: Fri, 26 Nov 2004 19:19:57 -0500
On 11/25/04 5:10 PM, "Graeme Gill" <email@hidden> wrote:
>> Tightening -by iteration- the match of your profile's output gridpoints,
>> however, should provide nothing but benefits. I don't know why they don't do
>> it!
>
> I'm not sure I follow you there. I certainly can't see any point of using
> iteration purely on an output profile. The forward profile has been
> determined by the sample points, and can be arbitrarily accurate when measured
> against those sample points, and the output profile is just the inverse of
> the forward profile, which once again, can be arbitrarily accurate
> when measured against the forward profile. Any meaningful iteration could
> be done purely in software. Of course the number and location of the sample
> points,
> together with the profile model and model fitting algorithms, will have a
> great
> influence on the real world output profile accuracy.
Graeme,
The target patches are defined in CMYK; thus the device->PCS is
straightforward, and can indeed be as accurate as the number and
distribution of the of patches allow. In fact, in many cases it's nothing
more than the formatting of the measurements originated by an evenly
distributed (in device values) target.
Furthermore, in profiles originated by various software packages I've
compared the device->PCS portion against the original target measurements,
and it's usually under 1 dE or close to zero.
But the pcs->device table has a grid defined by evenly spacing the PCS
coordinates; no actual patch has been read for that point. So the output
(CMYK) values in that table are calculated by interpolation of target
patches whose measurements yielded PCS values closest to the desired
gridpoint.
That interpolation is the one subject to error; because it's based on an
assumption (or model) which can be not totally accurate.
As a result, if you make a PCS (Lab) target file comprised of the gridpoints
(for example, Lab values 10,10,20 ; 10,10,30 ; 20,10,20 ; and so on) , pass
that target through the profile to convert to CMYK, print that CMYK in your
device, and measure, you won't get the desired 10,10,20 etc values.
You'll get close but not exact, because there lies the profile accuracy.
Of course I'm only talking about colorimetric intent for in-gamut colors,
which is what proofing is about.
But, alas, you can iterate. Based on that error measured, you can calculate
the correction needed to output (CMYK) values in the gridpoint and apply
such correction; and so on for the remaining gridpoints in this "Lab"
target. And of course (for the sake of consistency), interpolate those
correction vectors so as to apply to the in-between gridpoints that were not
part of the Lab target.
The more the device (CMYK) values in your output table result in actually
yielding the Lab values that define the gridpoint, the more accurate your
profile is, and the more accurate the linking (prebuilt or dynamic) with ANY
source profile.
In short, the more accurate your profile has become.
Why not?
-- Roberto Michelena
EOS S.A.
Lima, Peru
_______________________________________________
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