color rendering: a new approach · color rendering index (ra)) general preference rgb-low rgb-high...
TRANSCRIPT
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.© 2006 Rensselaer Polytechnic Institute. All rights reserved.
1
Color Rendering: A Tale of Two Metrics
Jean Paul Freyssinier, Mark S. Rea
CORM 2008
Sponsor: US Environmental Protection Agency
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.
2
Solid state lighting putting new demands on old metrics
There are different dimensions of color rendering:Naturalness, hue discrimination, vividness, preference, color naming accuracy
CRI used exclusively by lighting industryKnown limitations
Poor predictor of preference and color naming accuracy
Implications for energy savingsand standards
Basic issues
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.
3
www.lrc.rpi.edu/nlpip/publicationDetails.asp?id=901&type=2
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.
4
CRI and Color Naming Accuracy
CRI is not a good predictor of color naming accuracy.
90
91
92
93
94
95
96
97
98
99
100
60 70 80 90 100
CRIC
olo
r n
am
ing
accu
racy (
%) Xenon
MH+
MH
INC
CFL
0
10
20
30
40
50
60
70
80
90
100
0 20 40 60 80 100CRI
Co
lor
na
min
g a
ccu
racy(%
)
Xenon
MHP
MH
INC
CFL
HPS
LPS
0
10
20
30
40
50
60
70
80
90
100
0 20 40 60 80 100CRI
Co
lor
na
min
g a
ccu
racy(%
)
Xenon
MHP
MH
INC
CFL
HPS
LPS
An evaluation of the Hunt94 color appearance model under different light sources at low photopic to low mesopic light levelsLei Deng, Linsheng Chen, Mark S. Rea. 2005.Color Research & Application ; Volume 30, Issue 2, Date: April 2005, Pages: 107-117
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.
5
Skin Preference
RGB-lowRGB-high
amber white
phosphor LED
high power
incandescent
y = -0.6594x + 87.587
R2 = 0.441
0
20
40
60
80
100
0 20 40 60 80 100
Color rendering index (Ra)P
refe
rence r
ating
(%
)
General Preference
RGB-low
RGB-high
amber white
phosphor LED
high power
incandescent
y = -0.6215x + 115.03
R2 = 0.3448
0
20
40
60
80
100
0 20 40 60 80 100
Color rendering index (Ra)
Pre
fere
nce r
ating
(%
)CRI and Color Preference
Color Rendering Properties of LED Light SourcesNadarajah Narendran and Lei Deng. 2002. Solid State Lighting II: Proceedings of SPIE
CRI is not a good predictor of color preference.
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.
6
Basic issues
Bouma (1947): suggested that daylight is the best source because
can tell differences among colors (hue, lightness, and saturation)
can discriminate among subtle shades
can make objects appear natural
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.
7
Color Gamut Area
In general, the larger the gamut area, the more saturated the color samples are, and the easier it is to discriminate between them.
Adapted from Human Factors in Lighting - 2nd Edition Boyce, P. R. 2003. London; New York: Taylor & Francis.
Daylight
HPSMH
Fluor. Inc.
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.
8
90
91
92
93
94
95
96
97
98
99
100
30 40 50 60 70 80 90 100
Relative Gamut AreaC
olo
r n
am
ing
accu
racy (
%) Xenon
MH+
MH
INC
CFL
0
10
20
30
40
50
60
70
80
90
100
0 10 20 30 40 50 60 70 80 90 100
Relative Gamut Area
Xenon
MH+
MH
INC
CFL
HPS
LPS
Co
lor
Na
min
g A
ccu
racy (
%)
Gamut Area and Color Naming Accuracy
Color gamut area is a good predictor of color naming accuracy.
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.
9
General Preference
incandescent
amber white
phosphor LED
high
power
RGB-high
RGB-
low
y = 0.9032x - 5.331
R2 = 0.6137
0
20
40
60
80
100
0 20 40 60 80 100 120
Gamut area index (GAI)
Pre
fere
nce r
ating
(%
)
Skin Preference
incandescentamber white
phosphor LED
high
power
RGB-high
RGB-
low
y = 0.4682x + 0.9484
R2 = 0.1873
0
20
40
60
80
100
0 20 40 60 80 100 120
Gamut area index (GAI)P
refe
rence r
ating
(%
)
Gamut Area and Color Preference
Color Rendering Properties of LED Light SourcesNadarajah Narendran and Lei Deng. 2002. Solid State Lighting II: Proceedings of SPIE
Color gamut area can be a better predictor of color preference than CRI.
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.
10
Experimental Approach
Objective evaluationFarnsworth-Munsell 100 hue test
Subjective evaluationRestrict colors to red and blue
Restrict subjective responses to vividness and naturalness
Paired comparisons (sequential) for vividness and naturalness
Compare different light levels
Examine correlations of subjective and objective evaluations to CRI and GA
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.
11
Apparatus
Farnsworth-Munsell 100-hue color vision test
Collage of Cardinals and Blue Jays pictures
-
© 2008 Rensselaer Polytechnic Institute. All rights reserved.
12
Experiment setup
Daylight conditions
Electric light sources conditions