Ostwald and the

Theory of Colors
Jan J. Koenderin!
Physics & Astronomy, Universiteit Utrech"

Ostwalds Basic Contributions

to Colorimetry are:
the spectrum & the color circle (semichromes)
the mensuration of the color circle
the structure of the object color solid
the colorimetric color atlas
color & spectral signature
(I wi$ not discuss color harmony today)

The Spectrum
&
the Color Circle

Newton demonstrates the spectrum (George Romney)

Newtons spectrum had a resolution of only about 50-100nm

(but the colors were almost perfect!).

Newton saw 7(!) colors and fitted a musical scale to them.

Newton drew up a color circle (right), he probably emulated

Descartes representation of the musical octave (left).
He somehow managed to let the spectrum bite its own tail,
confusing the scientific world for over a century (till
Maxwells & Helmholtzs experiments).

B C G Y R

Newtons spectrum is
not complete:
Only the painters

color circle
contains all hues.
It remained for

R Ostwald to clear up

the relation between

the spectrum and

the color circle.

spectrum
inverted

spectrum

Newton thought he proved white

light to be a confused mixture of

homogeneous lights.

Goethe (Farbenlehre) showed that

complementary apertures yield

complementary images (Babinets

Principle) and produced the inverted

spectrum.

Neither the spectrum, nor the inverted spectrum are complete.

Goethe had trouble with green, Newton with purple.

Goethe experimented with

complementary apertures.
He came to consider colors as mixtures of edge colors
(Kantenfarben). This let to Schopenhauers notion
of colors as parts of daylight.

edge colors (Kantenfarben)

Goethes edge colors are obtained as the

cumulated spectra (starting at either spectrum
limit) of the spectrum of the illuminant (white
light, daylight). You see these colors when
you look at a light/dark edge through a prism.

The Goethe edge colors as

parts of daylight (cumulated
spectra from either spectrum
limit).

Schopenhauer (ber das Sehn

und die Farben) noticed that there
exist exactly two best bipartitions
of daylight.
The cut locus is at the antipodes
of the spectrum limits.

The cardinal colors as parts of daylight

Ostwald also thinks of the full colors as parts of daylight.

Indeed the semichromes are half of daylight. Because he
curved the spectrum the purples appear as natural parts.
Unlike Newton Ostwald retains the gap.

Helmholtz decomposition
0.8

0.6

0.4

monochromatic
component

0.2

0
0

0.5

1.5

spectral width

Ostwalds full colors have

maximum color content for a given
hue.
The band limits have to be at
complementary wavelengths in
order to obtain the maximum.

Ostwalds

semichromes

The relation between the color circle and the spectrum

is explained through Ostwalds construction of the
semichromes. The semichromes are also full colors.

The Mensuration
of the
Color Circle

The color circle is only a topological circle, without a

metric. One would like to fit it with a well tempered
scale. Ostwald pioneered a method to do this.

Ostwalds Principle of Internal

Symmetry is very simple: Define
bisection of line elements through
equal mixture.
Here the RGB color circle is

mensurated: Starting with the

primary colors R, G and B, one

obtains the secundary colors
C, M and Y. Going the other

way one obtains C, M and Y

from R, G and B. The whole

system is internally consistent.

Ostwalds Principle of
Internal Symmetry yields an
affinely invariant arc length
parameterization of the
color circle.

The first colorimetric

(numerical) calculations
were by Bouma in the
1940s (now forgotten).

500 nm
490 nm

510 nm
520 nm
530 nm
540 nm

480 nm

550 nm

560 nm

470 nm
450 nm
380 nm

570 nm

580 nm
590 nm
600 nm
620 nm
660 nm

Left an example of the mensurated color circle by

Ostwald. Right a mensurated color circle calculated
(using the CIE 1964 standard observer) from modern
data for average daylight (CIE D75).

Coloroid 48 step scale

490
500
480
470
460
450

Munsell 40 step scale

490
510
520
530
540
550
560

480

470

500
510
520
530
540
550
560

470
460
450
440
430
420
410
400
390
380
370

DIN 24 step scale

480
490
500
510
520
530
540
550

460
450

560
570

570

570

580
580
600

580
590

630620610 600 590

620
610

600

590

Psychological (Eye Measure) Scales

Empirically, the Ostwald mensuration predicts the
eye measure scales quite well. This is most remarkable
because it requires mere colorimetric (objective) data.
(Only judgments of equality, no absolute color
judgments!) Psychology is not involved.

The Structure of the

Object Color Solid

Of the many attemps to

systematize the object
colors before Runge (1834)
none came up with the
correct topology.
The modern theory of the
color solid is due to Erwin
Schrdinger (Ann.Physik
Lambert 1772

1920).

Philip
Otto
Runge

Runge came up with

the first color solid of
the correct topology.
(Farbenkugel, 1834)

The Schrdinger
color solid.

All object colors lie within a finite, convex volume. Its

boundary is smooth, except for two points: At the
black point the solid is tangent to the spectrum cone,
at the white point to the inverted spectrum cone.

off-white
black (monochromatic)
highpass edge color
lowpass edge color
bandpass (spectral)
bandgap (non-spectral)

The Schrdinger optimal colors make up the boundary

of the color solid. They are the brightest colors for any
given chromaticity (the best paints, containing no gray).
The most colorful optimal colors are the Ostwald full
colors.

Schrdingers color solid

mensurated via Ostwalds

principle of Internal Symmetry:

A fully objective framework.

Ostwald conceived of object colors as partitive

mixtures of white, black, and a full color. Placing the
gray axis perpendicular to the center of the
(full-)color circle, he arrived at a double cone.

example of
an Ostwald
atlas double
page
computed
for the RGB
colors

This atlas is
based on a
24 hue color
circle.
Notice the
idiosyncratic
hue names
assigned by
Ostwald.

white pole

ribbon of full colors

not a planar color circle!

black pole

In the Schrdinger color solid the Ostwald full colors

(Vollfarben) are at greatest distance from the gray axis.

The Ostwald atlas covers only part (A) of

white

the object colors (A+C+D) as derived by

Schrdinger (1920).

C
full color

A
D

ultimate color
(virtual)

By variation of the spectrum of

the illuminant region B is available.

The colors in region C are super-tints

(negative black content), those in region
D are super-shades (negative white
black

content).

100 %

50 %

0%
0

16

24

In terms of the ultimate colors the full colors themselves

can be analyzed in terms of their color, white and black
content! The best colors are blue and yellow, worst
are cyan and magenta.

The Colorimetric
Color Atlas

Albert Munsell (a painter)

worked on a color atlas
almost simultaneously with
Ostwald. (The two met.)
In the final instance, Munsell
reverted to eye measure to
arrive at a metric.

Albert Munsell

Thus the Munsell-system

stands well apart from
colorimetry proper.

Munsells concept of
a color tree is a
cylindrical coordinate
system based on
the gray axis.
There is no fundamental
limit to the distance the
branches may go.
This fully ignores the
colorimetric basis.

Construction of a page of the Ostwald atlas. The

samples can be generated by simple means (Ostwald
used the POMI, HASCH and inverted spectroscope).
This (in principle) involves NO EYE MEASURE.

Seegrn

Seegrn - Laubgrn
- Eisblau
Laubgrn

Ublau

Gelb - Laubgrn

Eisblau

Ublau - Eisblau

Gelb - Kress

Ublau - Veil

Kress

Rot - Kress

Veil
Rot - Veil

12

A sample from the Ostwald atlas:

Full color #12 (Ublau),
Ublau

12 850,
25<
5025,
25 25

50% color, 25% white, 25% black.

By far the best way

to memorize, select,
or indicate colors is
by way of their hue
and color, white and
black contents.
With some practice
one learns to see
these parameters.
Modern application: Color Pickers

In Ostwalds double cone the pages (full color-whiteblack triangles) are bound at the gray axis to a
periodic book (no need for front or back covers).

Although Ostwald used empirical methods

to find suitable pigments, his color atlas is
firmly founded on colorimetric principles

(NOT eye measure!) and can be computed

(and shown on the computer screen) from
standard colorimetric tables.

It is different in kind from Munsells system.

Color
&
Spectral Signature

Reflectance spectra of

natural materials (such as

skin) can be captured very

well through Ostwalds
scheme, at least for

colorimetric purposes.
700 nm

600

500

400

0.5

0.25

0
2.0

2.5

3.0 eV

Ostwald asked a physicist: What is the best recipe for

yellow paint? The answer was zero reflectance
thoughout the spectrum except at 580 nm.
Ostwald noticed that this will produce a BLACK paint!
Ostwald looked at spectral reflectance of actual good
paints and noticed that yellows reflect half of the
spectrum! There are no monochromatic object
colors.

black content

black

color content

color

white content

white
complementary wavelengths

Ostwald schematized spectral signatures in terms of

partitive mixtures of spectral reflectances: Black, white and
a full color. This cuts down the degrees of freedom to a
mere 3! For colorimetric purposes this suffices.
Ostwald proposed to cut down the spectrophotometry to
a determination of full color # and CWK contents.

Actual spectral signatures are not of the ideal

Ostwald type (although often not too different).
Thus the simple methods pushed by Ostwald will fail
to characterize spectral signature completely.
Ostwald was evidently aware of
this but failed to stress the point.
Although Ostwalds methods are
not exact they are often very
good approximations and quite
useful when applied wisely.

Some conclusions
Ostwald managed to clear up the relation
between the color circle and the spectrum
Ostwald managed to replace eye measure
with colorimetric denitions
Ostwalds color atlas may have a few aws, but
is a principled construction, in contradistinction
to Munsells mere eye measure result
Ostwalds spectroscopy (when understood
rightly!) has many useful applications.

What ever happened to

Ostwalds Color Science?
History has not been kind to Ostwald. His major achievements are
not acknowledged in the Anglosaxon literature and when
mentioned at all are misinterpreted. An embarrassment!
For example, Ostwalds color atlas is the only principled
construction of its kind available today. It is dierent in kind
from the Munsell system in that it obviates the need for eye
measure (psychology). It is perfectly suited to the computer
age. Yet this remains unrecognized, despite the need.
The current industry standard (CIE-Lab) is an awkward mix of
colorimetry and arbitrary denitions (loosely based on the
Munsell system), full of magical numbers and ad hoc functions.
Yet, commonly enough, people confuse it with science!

Thank you for your attention!

[email protected]