Re: colorsync-users digest, Vol 3 #1027 - 13 msgs
Re: colorsync-users digest, Vol 3 #1027 - 13 msgs
- Subject: Re: colorsync-users digest, Vol 3 #1027 - 13 msgs
- From: Marc Levine <email@hidden>
- Date: Thu, 30 Oct 2003 10:37:23 -0500
>
Is there any experimental evidence which indicates an optimal number
>
of samples? Is there variation in this number with respect to intent
>
(input, monitor, output, etc.)? Is there an over-arching theory which
>
would account for this discrepancy with intellectually "obvious"
>
results?
>
>
It seems that this is a usefully fertile realm of study and
>
experimentation: transitional smoothness can often be more desirable
>
than absolute color fidelity.
Dennis (et al),
There's 2 types of interpolation that are used when building a profile.
1) The "fitting of the cube". Profiles use a 3-dimensional cube to convert
Lab to device data (RGB, CMYK, etc). These cubes are comprised of nodes - or
fixed address points - from which most color information must be
interpolated. The number of nodes in a profile essentially dictates profile
size - profiles with more nodes are larger (in MB) than profiles with less
nodes. There are other factors as well, but I'll leave that for now. The
data that you measure from a profile target is used to fit the profile cube
with device data. As there are more cube points than actual patches (a
17x17x17 cube would have 4,913 nodes), interpolation is used to "guess" what
the node device values should be. More patches means more accurate node
values. Whether you measure more or less patches, it has no impact on the
number of nodes (that's actually set as an option when building a profile).
Profile smoothness is more a function of the number of nodes that anything
else and - with Monaco products - we see no difference in profile smoothness
using higher numbers of patches.
2) The second type of interpolation is "in-cube" interpolation. After you
get through mapping the cube with values, you've got to then interpolate the
device's "remaining colors". The more nodes you plot, the less you have to
interpolate. However, if a device is highly "irregular", it might be
difficult to smooth vales if there are too many nodes. Imagine a curve with
too many points. If many of the point are "off the curve", it will be
difficult to "smooth it out". Same thing with a profile. Monaco actually
gives you 4 cube resolutions to choose from - 9x9, 17x17, 21x21, and 33x33.
The higher the cube resolution, the greater the color accuracy (less
interpolation). The lower the resolution, the smoother the profile (more
interpolation). Keep in mind that it is the device's irregularities that are
the main culprit. If a device is well behaved (most inkjets are), higher
cube resolutions should have no impact on smoothness.
So what does this tell you? More patches should not affect smoothness. They
should only increase color accuracy. Higher cube resolution can increase
color accuracy even further, although it may compromise smoothness if the
device is irregular. If you are working with Monaco software, I would
encourage to contact me off line as the results that you (and Andrew) are
not typical of Monaco technologies. If you are using product from anybody
else, I would recommend switching to Monaco product. ;-)
Marc
--
Marc Levine
Sales Guy
Technical Guy
Monaco Systems / X-Rite, Inc.
www.monacosys.com / www.xrite.com
email@hidden
_______________________________________________
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.