[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Openexr-devel] Colour space
From: |
gary demos |
Subject: |
Re: [Openexr-devel] Colour space |
Date: |
Tue, 07 Sep 2004 11:57:14 -0700 |
User-agent: |
Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.4) Gecko/20030529 |
Chris,
Thanks for your comments. With respect to CIE 1931 XYZ (and xy), there
are some who agree with you,
that it is adequate, and there are some who feel that the errors are
significant. I have attached an interesting
example using the Judd/Vos 1978 correction to the x_bar, y_bar, and
z_bar functions. This example is
using LED color primaries. Note that the white/grey match is centered
in the inter-person variation point-cloud
using Judd/Vos1978, whereas CIE 1931 XYZ is at the lower-left of the cloud.
It is also worth mentioning that Judd/Vos 1978 only corrects cyan and
blue, and there are also know disparities
in the average in red, yellow, and green.
I have also seen an unpublished result testing Metal-Halide (UHP) lamps
vs Xenon in white matching, show
a similar systematic average error when using CIE 1931 XYZ. Note that
white matching is not likely to
be the most stressful match, so there are probably cases, using a few
spikey spectra where CIE 1931 XYZ
is further from the average viewer.
I am not sure that CIE 1931 XYZ is inadequate. However, I suspect that
this is the case.
Again, these opinions are my own, and do not represent any group.
-Gary Demos
Chris Cox wrote:
Yet neither RIMM nor ERIMM are linear (gamma 1.0) colorspaces. OpenEXR
files are supposed to be gamma 1.0. So you can't apply those
colorspaces as-is.
The problem with RIMM and ROMM, as well as with XYZ and most other
systems (e.g. Rec 709 HDTV) is
that the color primaries are specified in terms of CIE 1931 xy
chromaticity coordinates. This is also true for
the primaries in the OpenExr header.
That's not a problem - that's the standard, with good reason.
CIE 1931 xy chromaticity relies upon x_bar, y_bar, and z_bar spectral
matching functions, which are known to
have some errors in certain color regions (especially near cyan at
500nm, but also at other places in the spectrum).
Not much error according to the reasearch I've seen.
Yes, it could better account for women with a fourth primary sensor
and color deficient viewers.
I have been inquiring of color scientists concerning a) how much of
a problem are the errors in CIE 1931 XYZ,
and b) how much benefit is there from going to more than three color
primaries for the device-independent space.
a) not much
b) more than 3 primaries would present some mathematical difficulties
(N->1->N mappings anybody?)
Chris
_______________________________________________
Openexr-devel mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/openexr-devel
LED color match test.pdf
Description: Adobe PDF document