J. soc. cos. CHEM. 15, 79-85 (1964)
THE MEASUREMENT
OF SKIN COLOR,
SPECTROPHOTOMETRIC
TECHNIQUE
By WIL,4^M R.
M.D.* and FR^NKGRur, M.S.t
Presented
May 8, 1963, Semi-AnnualScientificMeetingof the SocietyoJ
CosmeticChemistsHeld in 5tointSponsorshipwith the American Medical
Association
Committee on Cosmetics
ABSTRACT
Early attempts to measure skin color used so-called "color appearance systems."
Because the eye is not an analytical instrument,
it
soon became evident that a physical system of measurement would
be required. It is shown that spectrophotometic measurements together with psychophysical correlations make it possible to specify
color in terms of its luminous reflectance (brightness), its dominant
wavelength (hue), and its excitation purity (saturation).
INTRODUCTION
The scienceof coloris concerned
with obtainingall the information
availableregardingthe productionof colorstimuli and the visual perception thereof. In the processof obtaining information, it borrowsfrom
physics,chemistry,biology,physiologyand psychology. One aspector
disciplineof the scienceof coloris colorimetry,which is the techniqueof
the measurement of color.
This is based on the fundamental
idea that a
relationshipcan be found betweenthe physical specificationsof color
stimuliandthe senseperceptions
that arisefrom them.
Very few attempts were made to measurehuman skin color until 1925.
At that time, in this country,Rowntreeand Brown developedand publisheda paperon their "Tintometerfor the Analysisof the Color of the
Skin" (1). They used somenine separatecolor scalesfor comparative
analysisand inter-relatedtheir systemwith the Munsell System (2). It
is of equalinterestthat oneof the mostrecentcolorappearancesystemsto
be developedis that by Dr. H. RichardHellstrom(of the Veteran'sHospital
in Pittsburgh) who had long been concernedwith the nonuniformityof
colornamesusedfor the descriptionof grosspathologicalspecimensboth
* Senior Clinical Instructor in Medicine (Dermatology), The University of Rochester
Schoolof Medicine and Dentistry, Rochester,N. ¾.
t Eastman Kodak Co., Rochester,N. ¾.
79
80
JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS
in operatingand dissectingroomsand in the generalmedicalliterature.
This "Method of ColorDescriptionfor Use in GrossPathology,"an adaptation of The ISCC-NBS Method of DesignatingColors,(3) is expectedto
be of somevalue if it is acceptedby pathologists. The principaldrawback
to the use of sucha simplecolor chart is the preconceived
notion of different observers of what each color should be called.
Both of thesetechniques
of measuringand specifyingskincoloror color
of biologicalmaterialsare mentionedherebecausethey are fundamentally
basedon a "colorappearance
system." In this type of systemthe colors
are selectedand arrangedwith respectto appearance. The relationships
involvedarepsychological
rather than physicalor psycho-physical.
It is alsogenerallyagreedthat it is impossibleto representan equally
spacedcolor system in three-dimensionalspace. In addition, the best
arrangementby psychological
attributesis not the bestfor obtainingequal
spacing. The productionof physicallyreproduciblestandardsfor colors
of the spaceis usuallyaccomplished
by resortingto C.I.E. (4) tristimulus
and reflectancevaluesfor samplecoloror colorantcombinations;however,
suchcombinations
do not give a seriesof colorsconformingto equalspacing
by psychological
attributesin three-dimensional
space.
The earliestattemptsto measureskin colorusinga spectrophotometer,
at least in this country, were made by Sheard and Brown (5). Their
work was publishedin 1926. A few years later this original work was
expandedby Sheardand Brunsting,and publishedin 1929 in a seriesof
three papers (6). Their work was excellentwhen it is realizedthat the
spectrophotometer
they usedwas crude, and the newertristimulusvalues
were not evenagreedupon at that time.
The final project putting spectrophotometry
on a soundbasisin the
measurementof skin colorwas that by Edwards and Duntley in 1939 (7).
This work establishedmost of the skin colorants. Moreover, the Hardy
Spectrophotometer,
a really fine instrument,was utilized for the measurements.
STATEMENT OF PROBLEM
In this paper, an attempt will be to specifyhumanskin colorin terms
of a definiteset of physicaloperations. This is not unusualwhen it is
realizedthat any definitionof a quantity is only a statementof the proceduresby whichthe quantity is appraised.
First of all, the reflectanceof the skin will dependon the following
factors(8):
(1) The surfacelayersof the epidermis,
(2) The amountanddistributionof the pigments,
MEASUREMENT
OF SKIN
COLOR
81
(3) The over-allturbidity of the skin (duechieflyto the melaningran.
ulesand cellularconstituents),and
(4) The fillingof thesubpapillary
plexus.
A second feature of the reflectance is the fact that a cross section of
human skin, as observedfrom the surfacedownward,shouldbe considered
as multilayered. As the incident flux within the spectrophotometer
strikeseachlayer: A portionis reflected;a portionis scattered;a portion
is absorbed;and a portion is transmittedon downwardto the next layer.
The samephenomenaoccur at that layer, and so forth until the energy
of the incidentbeamis dissipated. The total spectralreflectancemeasured
by the spectrophotometer
contains,therefore,the absorptionbandsdue
to the variouspigmentswithin the layersof the skin.
RESULTS AND DISCUSSION'
The followingmeasurements
of total spectralreflectanceweremadewith
the General Electric Spectrophotometer:Figure 1 showsan averagereflectancecurve of the skin taken from 10 different white subjectsof both
sexesand with various complexions. The skin area measuredwas the
malar regionof the left cheek.. The general characteristicof the curve
is its generalslopedownwardfrom right to left. In other words,skin reflectsmore spectralred and yellow but lessviolet, blue and green. This
gradualslopecan be shownto be due to the pigmentmelanin. There is
a strongabsorptionbandat approximately415 mtadue to the hemoglobins
--chiefly oxyhemoglobin.This dip is somewhatless than might be expected becauseof marked scatter in this regiondue to melanin granules
and the other cellularconstituents. Proceedingagain to the right on the
curve,there is a slopethat extendsfrom 465 to about510 m**. This slope
has a maximal absorptionband due to the carotenoidsat approximately
480 mu. Continuingstill further to the right on the curve, there are two
dipsor absorptionbandslocatedat 542 and 575 mu, due to oxyhemoglobin.
There is no recognizable
absorptionbanddue to reducedhemoglobinnoted
in the normal reflectance curve.
Spectralreflectancecurvesare, in the beginning,rather confusingwhen
they are studiedby thosenot acquaintedwith them. However, they
providevastly moreinformationthan can be obtainedby visualexamination of the color of the object in question. Moreover, the information
that can be obtainedfrom the curve is essentialin the solvingof many
colorproblems.
How are spectrophotometric
data used to developa color specification
for human skin? It has been known for well over a hundred years that
very nearly all colorsencounteredin nature can be matched by merely
adjustingthe intensities
of the red,greenandbluecomponents
of a mixture
of coloredlight. Sucha mixturemay bereadilyproducedby threelanterns
82
JOURNAL OF THE SOCIETY OF COSMETIC
CHEMISTS
arrangedto project three partially overlappingspots--a red, a greenand
a blue on a white screen. Where the greenand red overlap,the mixture
appearsyellow. Where the blue and red overlap, the mixture appears
magenta. Where the blue and greenoverlap,the mixture appearscyan.
Whereall threebeamsor spotsoverlap,themixtureappears
white.
60
i
i
_.m30
o• 2o
I VioletBlue GreenYellow Red
400
500
600
700
Wovelength
(mp.)
Figure 1.--Total spectralreflectancecurve--averagewhite skin,
Thus the data of color vision on which color measurement must be
basedare obtainedby systematicexperimentsin which many colorsare
matchedvisually. The intensitiesof the red (-•), green(•), and blue (2)
components
necessaryto match equal amountsof energyat eachwavelengthof the spectrumcan be plotted as curves,shownin Fig. 2. These
are called colormixture functions. They have beenstandardizedto very
great accuracyby the InternationalCommission
on Illuminationand by
the American Standards Association.
Using I.C.I. Illuminant C (whoseenergydistributionover the visible
spectrumis known) as the representativesourceof averagedaylight,it
is possible
to obtainthe spectraldistributionof the light reflectedfromthe
skin (9). This is accomplished
by multiplyingper cent reflectancetimes
energyat eachwavelengthof the spectrum. The curve so obtainedcan
then be multiplied,wavelengthby wavelength,by eachof the threecolor
mixture functionsto obtain the fundamentalspecificationsof the color
of the skin (10). From the averagereflectance
curveshownin Fig. 1, the
tristimulus values would be:
X
= 38.0
¾ = 29.2
Z = 32.9
MEASUREMENT
OF SKIN COLOR
83
2.½
Color
Mixture
Functions
1o0
0.0
•i
Wavelength
in m•
6o•
700
Figure 2.--CIE tristimulusvalues•, 3, and • for the spectrumcolors.
Becausethe • functionchosenas one of the primariescorresponds
exactlyto theso-called
visibilityfunctionfor a normaleye,the relativebrightnessof the skinsampleis givenas29.2per cent,whichwouldbeequivalent
to its Y function.
84
JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS
.BO
.75
.70
.6.5
.5.3
.StJ
.4.3
.40
.33
30
2.;
.05
o
.03
.10
.15
.20
.25
.30
.35
.l/J
.t.;
..;n
.5.;
.60
.65
.70
.75
Figure3.--Chromaticity
diagramfromaveragetotalspectralreflectance
curve.
Thequalityof theskincolorisobtained
by defining
threenewquantities,
the trichromaticcoefficients. In thisparticularcase:
x = 0.3786
y = 0.3374
z = 0.2840
Plottingthe x andy trichromatic
coefficients
on a "ChromaticityDia-
MEASUREMENT
OF SKIN
COLOR
85
gram," (Fig. 3) it can be determinedthat the averagereflectanceof 10
samplesof white skin as shownin Fig. 1 has a dominantwavelengthof
593 mtz (yellowishhue) and an excitationpurity of 24.0 per cent. Thus
you have a completecolor specificationfrom the spectrophotometric
findings.
Beforeclosing,it shouldbe pointedout that the valuesas given do not
include any contribution from the fluorescenceof the skin. Under ordinary circumstances
this presentsno problem. However, under certain
circumstances,
it may be necessaryor desirableto measureskin fluorescence
and its effect on over-all skin color.
Grum and Wightman (11) have presenteda paper on the techniqueof
measuringthis fluorescence
by modifyinga spectrophotometer
in such a
way that it will recordfrom the sampleat eachwavelengththe "complete
radiation." Completeor total radiationis the sumof the fluorescence
and
reflectance
from the sample.
SUMMARY
Colorimetryis the techniquefor measuringthe colorof objectsor surfaces. The earliestattemptsto measureskin colorwere by using"color
appearancesystems." Becausethe eye is not an analyticalinstrument,it
soonbecameevident that a physicalsystemof measurementwould be required. The spectrophotometer
providedthis approach. Throughpsychophysicalcorrelations,it becamepossibleto use spectrophotometric
data to specifycolorin terms of its luminousreflectance(brightness),its
dominantwavelength(hue),andits excitationpurity (saturation). Special
procedures
are requiredwhenthe effectof fluorescence
is to be includedin
measuringskin color.
(ReceivedJune 10, 1963)
REFERENCES
(1) L. G. Rowntree and G. E. Brown, `4m. •. Med. Sci., 170, 341 (1925).
(2) A. Munsell,.4 ColorNotation,6th Edition, MunsellColorCo., Baltimore,Md., 1923.
(3) NationalBureauof Standards,
ISCC-NBSMethodof DesignatingColors,NBS Circular
553, GovernmentPrinting Office,WashingtonD.C., 1955.
(4) Commission
Internationalede l'Eclairage,Proceedings
of theEighthSession,
Cambridge,
England,1932,pp. 19-29.
(5) C. Sheardand G. E. Brown,.4rch.Intern. Med., 38, 816 (1962).
(6) C. SheardandL. A. Brunsting,•. ClinInvest.,7, 559 (1929); Idem., •e.Clin.Invest.,7,
575 (1929); Idem., •. Clin. Invest.,7, 593 (1929).
(7) E. A. Edwardsand S. Q. Duntley, .4m. •..4nat., 65, 1 (1939).
(8) W. R. Buckleyand F. Grum, .4.M..4..4rch. Dermatol.,83, 249 (1961).
(9) A. Hardy, et al., Handbookof Colorimetry,
Preparedby the Staff of the ColorMeasurement Laboratory of Massachusetts
Institute of Technology,The TechnologyPress,
Cambridge, 1936.
(10) D. L. MacAdam, T.4PPI, 38, 78 (1955).
(11) F. Grum andT. Wightman,Ibid., 43, 400 (1960).
86
JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS
NOTICE
S.C.C.
TO
MEMBERS
Many membersof the Societyhave requestedthat a completerosterof their fellow-members
be madeavailable.
Such a membershiplist is now in preparationand, it is
hoped,will be publishedin oneof the 1964issuesof the Journal.
To be sure your own listing will be accurate,pleaseadvise the New York office,as soonas possible,of any changes
in your address. Pleasesendthis informationto:
Society of Cosmetic Chemists
Mrs. Eunice Thomas Miner,
Administrative
Assistant
2 East 63rd Street
New York 21, New York