Re: Illuminant change
Re: Illuminant change
- Subject: Re: Illuminant change
- From: Robin Myers <email@hidden>
- Date: Wed, 04 Apr 2001 08:52:16 -0700
- Organization: Robin Myers Imaging
Dan,
If you change the illuminant, the XYZ and L*a*b* values will change. The XYZ
values are calculated from a product of the reflectance, the illumination and
the observer. The L*a*b* values are then calculated from the XYZ values.
Therefore, as you change XYZ you can expect the L*a*b* values to change. It is
best to use spectral data to calculate the L*a*b* results of illuminant changes.
If you change the illuminant and the XYZ values change, but the L*a*b* values
do not change, then something is wrong. I suspect you had the rendering intent
set to perceptual or relative colorimetric, which will try to keep the L*a*b*
values constant. This is something some color engineers have done to make life
easy for them but this is wrong. The L*a*b* space was not designed for
calculating appearance, especially when changing illuminants. It was designed
only for calculating color differences and has even been found wanting in this
regard (hence the modifications to Delta-E(ab) such as CMC and CIE Delta-E(ab)
1994, which attempt to make the differences more uniform).
The most recent work on an appearance model, which will predict the change in
perceived color with illuminant changes, is the CIECAM 97 work. It is still in
its infancy, changes often, and has not been put through extremely rigorous
testing, yet. So stay tuned on this issue.
Creating color matches tuned for the display illumination is optimal, but
often impractical. It is very difficult to get users to understand that once a
print is tuned to a particular lighting, they must adhere to this requirement.
The user may ask for a match made for tungsten lighting which may be displayed
under cool white fluorescents. When the image does not look true to the
original, who gets the blame? Look at all the problems Epson is having with
the 2000P and its metamerism issues.
This is what the standards for matching and viewing were supposed to address.
In the U.S. the standard is D50, equivalent to direct sunlight viewing. In
Europe, the standard is D65, equivalent to indirect sunlight viewing. Which is
best? You must decide. But remember, the exact match is only valid for the
standard viewing conditions.
Can prints be made that do appear similar under different lighting, yes,
but... It is mainly a function of the colorants used, the substrate materials,
the display environment (e.g. surrounding colors) and the illumination (there
are other factors but they tend to be lower order effects). As the
illumination changes, the goals are to have prints that maintain their basic
color balance (e.g. do not acquire a color cast), memory colors remain
somewhat similar (e.g. sky stays blue, grass green, flesh tones maintain
healthy appearance) and the tonality of the image is maintained. Most printing
systems achieve this. The printing industry has worked on this issue for a
long time, specifically to address the metamerism issue and has selected
colorants and materials that minimize problems with illumination changes. As
an example, take a magazine page into different lighting. The images and
colors maintain their basic relationships even though they may be viewed under
tungsten, fluorescent, natural light, or combinations of these. The exact
color values do change, but the relationship to the whole image is maintained
within the tolerance of our acceptability.
Let me know if this helps or hinders and I will attempt to clarify these issues.
Robin Myers
Robin Myers Imaging
email@hidden