Re: [Gimp-developer] Soft proofing and the GIMP Display Filters and Color Management settings
- From: Elle Stone <ellestone ninedegreesbelow com>
- To: Gimp-developer <gimp-developer-list gnome org>
- Subject: Re: [Gimp-developer] Soft proofing and the GIMP Display Filters and Color Management settings
- Date: Wed, 12 Mar 2014 17:27:20 -0400
On 03/11/2014 02:39 PM, Omari Stephens wrote:
That said, I think your question touches on something that I'm pretty
ignorant about, which is how color profiles deal with different numbers
and types of channels.
How monitors can use more than three channels (some already do) to make
colors is easy to visualize:
http://ninedegreesbelow.com/photography/all-the-colors.html#some
Some cameras also use more than three colors (sensor lens cap colors) to
capture colors.
Hopefully the printer people will correct me if I'm speaking nonsense
here. CMYK printer profiles have four channels because ink produces
color subtractively, but not perfectly, as inks are not as "narrow pass
reflective" as one might like. So using C+M+Y to make black produces a
muddy black and uses a lot of ink, which is sloppy to print. So the
fourth color is black.
More than four colors of ink gives smoother color reproduction and also
may extend the available color gamut, depending on the inks. The
corresponding ICC profile is a Lookup Table profile, which basically
says "r% ink-1 + s% ink-2 + t% ink-3 + u% ink-4 + . . . + z% ink-n"
(where r, s, t, u, . . . z are arbitrary percentages) equals a
particular location in the CIELAB reference color space, for all
possible combinations of various percentages of the n available inks.
I know there is a thing called a GRAY profile,
but I have no idea what makes it special or different than a standard
profile. Is it basically an RGB profile with R==G==B?
Yes in effect, no as far as the actual grayscale profile goes. A
grayscale profile only has a single channel that holds luminance
information in the form of a curve that goes from 0 to max white. Any
RGB matrix profile with an equivalent tone reproduction curve will
"match" the grayscale profile as long as R=G=B for all pixels in the RGB
image. Each channel in the RGB image will match the single channel in
the grayscale image.
I've never seen a LUT grayscale profile. But for matrix grayscale
profiles, the only information needed to define the profile is the white
point, the black point (which is zero for all the matrix Grayscale
profiles I've ever seen), and the tone reproduction curve (aka tone
response curve). RGB matrix profiles also need red, green, and blue
colorant tags.
From a practical point of view, less disk storage is required for a
grayscale image, compared to an RGB image with R=G=B for all three channels.
Does GIMP uses less RAM and/or CPU power to process a grayscale image,
compared to an RGB image where R=G=B?
In the New GEGL World, converting between different channel layouts is
going to be a reality, and we should at least put _some_ thought into
what that means for color management. Of course, this is way out of my
depth, and I have no idea.
I'm also curious as to what gegl n-channel editing might be like. Soft
proofing to an n-channel printer is a one use case for n-channel
editing, when the goal is to convert to the n-channel ICC profile and
tweak the channels while soft proofing. Hopefully again the printer
people will correct me if I'm speaking nonsense.
Dan Margulis gives examples of image editing in an artificial CMYK
matrix color space, requiring four channels.
Would there be a use case for editing in n-space (as opposed to soft
proofing to an n-space output profile), where n is greater than 4?
LCMS2 already accomodates n-channel ICC profile conversions. The GIMP
lcms.c plugin doesn't, but support could be added.
Elle
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]