Re: Gradients
Re: Gradients
- Subject: Re: Gradients
- From: Armand Rosenberg <email@hidden>
- Date: Fri, 31 Oct 2003 12:55:00 -0500
I agree with CD and do not follow the penguin and egg arguments of others.
The question was about smooth gradients in printing I believe.
The point here is not so much to get a minutely accurate
representation of an object (such as a printer gamut), but rather a
smoothed representation of it (while still as accurate as possible).
If you have an object with some small features (call them wrinkles)
the more measurements you take the better you will reproduce those
wrinkles. But this could well lead to gradients that are not so
smooth if they happen to involve one of those wrinkles. If instead
your intent is to smooth out those wrinkles (eg, for obtaining
smoother gradients) then certainly ONE valid way to do it is to take
fewer measurements. Somewhat equivalently, the software that builds
the profile can have a smoothing parameter (and some do). In fact,
such a parameter (or even an entire algorithm) is better in
principle, but it would require several things: (1) the software
vendor has to implement those features, (2) the software vendor has
to make those features accessible (in all ways) to the user, and (3)
the user has to have enough smarts and interest to use them. Since
these are tough requirements to meet, the simplest approach that
requires no effort on anyone's part while often achieving similar
results is to just take fewer measurements. The danger is that some
oddly shaped objects will be misreprented, but as long as the shapes
are well behaved (in all senses) this is a perfectly valid way of
smoothing out the wrinkles (for smoother gradients).
Armand
At 6:49 PM -0800 10/30/03, email@hidden wrote:
Marc's explanation follows the "makes sense" criteria to me, although
this is not my field.
If one is to measure an egg, or a penguin or any object, the more data
points that are obtained, the better definition of the shape and size of
the object. That is still true even if the individual measurements are
noisy. Actually, it ought to work specially well if the measurements are
a little noisy since more data points will be averaged to provide a
better definition of the object.
Why there apparently are different results in the field is not clear to
me. Do some profiling engines choke on too much data? Do they fail to
internally average the data to provide a smoothed fit or do they take
each point as an inflexible data point?
CD's comment about nailing a carpet vs. stretching it a clever
explanation for the output section of the profile although I do not see
it applying to the input side of the system.
So many question, so little I know on the subject.
Bertho
_______________________________________________
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.