|
|
Instruments can measure SPECTRA, but NOT color, until the numbers are related to human vision and light sources. Here we discuss the history and the state of the art. We provide links to important existing data sets, and present our wish list for more and better light source, illuminant, and observer data!
Return to | Internet Data Base | COLORING Index | or visit | Bill Dawes' Home Page |
Index
- Introduction to color measurement
- What's not here -- can you help?
- CIE Observers -- narrow and wide field
- CIE Illuminant A (tungsten) and D (daylight) illuminants
- The above, abridged observer and light source data tables and plots
- MH and fluorescent lamps
Introduction
COLORIMETRY IS THE MEASURING OF COLOR, to permit quantifying of perceived differences between items. Putting useful numbers on the colors of lights requires knowledge of the spectral response of the human observer's eye. Then, using the radiant spectra of lights, we can calculate color differences between lights transmitted through filters, or reflected from objects.
UNIFORM COLOR SPACE SYSTEMS are used to calculate the numeric differences between items. When there are dozens of ways to do the calculation, it safe to assume that none of them are very good. 'New, improved' systems appear every few years. The goal is to provide spheroids in color space, oriented so that one unit of color difference is perceived as equal in any direction. Most useful systems provide hue, value, and chroma vectors. Some of the better ones are described by Roberts and Ford in "Colour Space Conversions."
EARLY ATTEMPTS to relate instrument values to human perception seemed to flounder around creating new scales. The work by Dorothy Nickerson and Deanne Judd, creating spacing in the Munsell colors, was a significant advance. Finally, researchers agreed to use the CIE xyY color space, and simply modify the arithmetic.
RICHARD HUNTER, an instrument maker, created the first widely used Hunter delta E for color differences. David Macadam, working with observations of lighted fields by his tireless co-worker Perley G. Nutting, created ellipsoids in CIE space that described 'just noticeable differences.' (Young Perley, a grad student at the time, is believed to be the son of Perley G. Nutting, who founded of the Optical Society of America in 1916. I know nothing about our First Standard Observer beyond this. If you know more, please tell me -- he could become a model for technical grunts like me...)
Like so much theoretical work, the complex math involved was rarely used until digital computers became common tools, and the Macadam unit (FMC II dE) was universally adopted in color matching programs. Some still use it today, though better systems have been developed.
PRACTICAL COLORIMETRISTS weren't terribly happy with Macadam units. Sometimes, especially in bright colors, visually acceptable color matches would differ from the target color by four or six units. In 1976, a group of workers got the CIE to approve the CIE dE* (now abbreviated CIE76 dE), where one unit might be equal to several Macadams. Experimental work showed that this system was better with bright, high chroma colors. Colorimetrists were happy to change to the new system, largely because it was more comfortable to tell customers that the next lot of yellow paint was only .9 delta E from target, rather than four Macadams. CIE76 dE remains in wide use today, and almost all instruments put out values in this system.
BETTER SYSTEMS have recently been developed. The proprietary Datacolor system enjoyed some popularity in Europe, but was only used on that company's instruments. Then Roderick McDonald of the Coates company created the CMC difference, working with a huge body of experimental data from quality judgements on close matches of thread samples. A similar, contemporary system is called the BFD. And the CIE has adopted a close mimic of the CMC unit, and named it the CIE94 dE. The CIE94 unit is said to be easier to calculate, though that advantage seems trivial now that fast computers do all the hard work. However, McDonald endorses the switch. It's also been adopted by the AATCC.
BOTH THE CMC AND CIE94 SYSTEMS allow for adjustment factors. McDonald recognized that dull finish threads and fabrics would need differently shaped acceptablility spheroids than glossy paints, plastics, and ceramics. Workers in these materials found that they should put more weight on the light-dark vector. So the CMC and CIE94 systems anchor the orientation of the spheroids in color space, but they allow that future workers may justify changes in the initially suggested dimensions.
What's not here
I'm still looking for these data:
The Deluxe Cool White, Spectralume, any important fluorescent lamps. Daylight from dawn to dusk in Phoenix AZ and Newark NJ. The deviant observer (not a voyeur -- but the extreme of 'normal' observers) Anomalous, or color-blind observers -- particularly the Red-Green Caucasian male, of which I am one. Aged observers with yellowed lenses -- again, I'm included here. If you're as lucky as me, you'll eventually achieve a similar status some day. Improvements on the CIE Standard Observers.
COLOR DIFERENCE PROGRAMS TODAY that use the CIE observers with radiometric data from the new lamps are putting out crude approximations, and few users of the data are aware of this serious flaw. William Thornton has demonstrated that the CIE observers which were satisfactory for daylight and tungsten lamps fail to describe what happens with metal halide and fluorescent light sources. The coloring community is anxiously awaiting publication of improved weighting functions, based on real measurements rather than assumptions.
Ways to fine-tune CMF data to match your instrument's band-pass and wave shape. Software to read our tables and do color differences with your submitted spectra. A 'universal data format' for spectra and color data for the future Net Aware spectrophotometer. JCAMP-DX? ASTM? A blend, or both? Can YOU help with any of these goals?
CIE Standard Observers
The CIE (Commission Internationale d'Eclairage is French, means International Commission on Illumination) has a Home Page at http://www.hike.te.chiba-u.ac.jp/ikeda/CIE/home.html, or check their README file for descriptions of the following files:
You'll find the same data and more, in a beautiful format from the UCSD Color and Vision Research Labs page http://cvision.ucsd.edu/index.htm, maintained by Andrew Stockman and Lindsay Sharpe. Here you'll find CIE and other data at 1 and 5 nm increments, along with anamalous observers and other data of interest to vision researchers.
CIE Standard Illuminants
- Illuminant A, equivalent to a tungsten lamp.
- Illuminant D65, like an overcast daylight.
- Approximate daylight, from 4,000K to 25,000K. Components S0, S1, and S2, and a guide to calculations by Danny Rich.
Abridged CIE Data
- If you would prefer an abridged version of the above data more suitable for spreadsheets (at 10nm increments from 360 to 740nm), consider
- Table of lights and observers. I made the table using MS-Excel: no guarantees! The link includes cute plots.
MH and Fluorescent Lamp Spectra
- Four radiant spectra from Mitsubishi Electric: a metal halide lamp, and daylight, "moon white", and nominal 4000K fluorescent lamps. Table formatted data, with plots.
the World.
