Re: Monaco 4.0 editor
Re: Monaco 4.0 editor
- Subject: Re: Monaco 4.0 editor
- From: Graeme Gill <email@hidden>
- Date: Tue, 11 Jun 2002 13:36:53 +1000
Roberto Michelena wrote:
>
> If a profile package doesn't give you accurate greys,
>
> it's simply not doing its job.
>
>
The profile package and the measuring instrument are together in this.
>
Given the basic prerequisites, such as printer repeatability and
>
standardized viewing conditions, technology is still not perfect:
>
1) most instruments are confused by the "over100%" reflectance -at different
>
wavelenghts- returned by fluorescent inks and papers. Some just don't read
>
over a certain reflectance. Sensor saturation? electronic clipping? who
>
knows, the fact is it happens.
Hmm, - I haven't noticed any problems with instruments like the
Xrite DTP41 or the Gretag Spectrolino. They seem to handle over 100%
reflection quite well. As for fluorescence in general, luckily few
high quality, long life inks seem to use fluorescent dyes, and the
fluorescence often present in the paper stock can be compensated
for reasonably well in software (and no, I'm not talking about
the so called "UV filters" that are sometimes fitted to spectrometers).
The trouble with viewing conditions is that too often the viewing
conditions are ignored in the creation of color profiles, rather
than accounted for. The fix for this sloppiness is then the laborious
hand tweaking of the profiles.
>
2) using a discrete, and relatively -to the color space covered- small,
>
number of patches, the profiling app might not catch a quirk caused by a
>
quirky printer driver at a certain point. Not only crossovers in gray, but
>
also inverted ramps, etc.
Most profiling packages are incredibly inefficient at characterizing the
device space. The IT8.7/2 for instance is terrible. It uses a cubic lattice
evenly spaced in device space. Cubic packing is about 30% less efficient
at filling 3 dimensional space than simplex packing (and worse again in
4 dimensions !), while the regularity of the test points in device space
is a very poor choice for creating robust, detailed per-device linearisation
curves. Spacing test points uniformly in device space is a very poor
choice, since characterization errors are seen in perceptual space.
Ideally test points should be spaced more closely at high curvature of
the response space, and less in regions of low curvature. Overall something
like an IT8.7/2 is probably 50% as efficient as it should be.
>
Not to mention all the subjective and endlessly debatable things, such as
>
"how average the standard observer is?" and "my blue is not your blue".
This issue doesn't go away with tweaking profiles, in fact it may make
it worse, since you will be adjusting to suite one single observer,
who may not be near the average of all observers. It does inherently
take account of the viewing conditions though.
There are other observer curves out there too, based on
more recent research than the "1931 standard observer".
>
So given the imperfect technology as (1) and (2), yes, profiles usually need
>
manual tuning if you want the absolute best match.
I don't agree. A skillful tweaker may be able to make a single
test image look slightly better that what a good profiling package
can achieve, but they will only be playing with one "slice" of
the color space exposed by that image. As soon as you extend
the task to the full 3d colorspace (or even 4d space if you
want to tweak a "through black" profile), I doubt the task is
feasible in any reasonable amount of time. Multiply by media, modes,
number of machines and differing levels of skill, and the situation
soon gets out of control.
>
And as Johan said, from then on you can go doing great prints time after
>
time... And with a good profiling package and/or rip, linearization is all
>
it'll take to keep them accurate over time.
Sure, but this is true however you arrive at a good profile. But every
time you change the printer mode or want to use a different stock (or even
want to setup for a different source target, if you were compensating
for source and destination profile errors by tweaking the destination
profile only), then it is back to square one, and lots of hand tweaking.
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.