CATEGORICAL COLOR RENDEING OF LED LIGHT SOURCES H. Yaguchi, N. Endoh, T. Moriyama and S. Shioiri CIE...

CATEGORICAL COLOR RENDEING OF LED LIGHT

SOURCES

H. Yaguchi, N. Endoh, T. Moriyama and S. Shioiri

CIE Expert Symposium on LED Light SourcesJune 7, 2004, Tokyo

INTRRODUCTION

• Energy efficiency and color rendering– CIE CRI is based on color difference

• Importance of color name– To communicate color information in every day life.

– To categorize objects and recognize them.– Color name does not depend on viewing condition.

• Color rendering based on color categorization

PURPOSE

• To evaluate color rendering of LED light sources to take account of color categorization.

• The Subjective method by categorical color naming experiment.

• The Objective methods using the CIE color rendering index and the categorical color rendering index.

EXPERIMENT

Color samples

• 292 Munsell color chips• V: 2, 4, 6, 8• C: 0, 2, 4, 6,• • •• H: 5R, 10R, 5YR, 10YR, 5R,• • •

Spectral power distributions of light

sources

00.20.40.60.81

300400500600700800wavelengthÅinmÅjrelative spectral irradiance

D65LED(7500K)LED(5100K)LED(4000K)LED(2800K)

Categorical Color Naming

• Sort samples into 11 basic color categories: – red (aka), gree (midori), yellow (ki), blue (ao), orange (daidai), pink (momo), purple (murasaki), brown (cha), white (shiro), gray (hai), and black (kuro).

• Sorting was repeated three times.• Select samples sorted into the same category for all three trials.

Subjects

• 9 subjects• Age from 21 to 55• Normal color vision checked by the ND-100 hue test.

Color samples in the Munsell hue circle, named

consistently

R

P Y

B G

R

P Y

B G

R

P Y

B G

R

P Y

B G

R

P Y

B G

Result

Observer Kn, v=6

LED(7500K)　　　　　　 (5100K)　　　　　　 (4000K)　　　　　　 (2800K)　

D65

• Munsell color circles

Same color samplesare named different color

Percent Correct

• ND65: number of samples sorted into a color category consistently for three trials under the D65 simulator.

• Ns: number of samples, under test light source, sorted into the same color category as in the case of the D65 simulator.

• Nd: number of samples, under test light source, sorted into the different color category from the case of the D65 simulator.

• Percent correct = 100(Ns - Nd)/ ND65

Percent Correct for 8 color names

0

20

40

60

80

100

Red PinkOrangeYellowGreenBlue

PurpleBrown

Color Name

Percentage of correct (%) LED7500K

LED5100K

LED4000K

LED2800K

CIE color appearance model (CIECAM97s)

CIECAM97sINPUTLuminance of adapting fieldTristimulus values of sampleTristimulus values of whiteParameters for viewing conditionLAX, Y, ZXw, Yw, ZwOUTPUTHue, hLightness, JBrightness, QSaturation, sChroma, CColorfulness, M

Color map in CIECAM97s75<J 55<J<75

35<J<55 J<35

How to calculate

●CCRI for each test sample

● Averaged CCRI

CCRIi=100•Si∩ StSt

CCRI=( CCRIi)i=1

8

∑ /8

Categorical color rendering index (CCRI)

×

×

×

×

Boundary of color samples under the test light source “St”

Boundary of color samples under D65

×

×

×

× Cmax

Hmin

Hmax

Cmin

Chroma

Hu

e an

gle

Boundary of Color name “Si”

CCRI of 5 light sources

0

20

40

60

80

100

120

Red Pink Orange Yellow Green Blue Purple

Color name (i)

CCRIi

D65 Simulator

LED7500

LED5100

LED4000

LED2800

Ri, CCRIi and percent correctD65 LED75 LED510 LED40 LED28

redR9 83 52 -1 14 -5

CCRIred 97 88 90 91 83

PCred 100 73 83 73 63

yellow

R10 90 94 74 94 91

CCRIyellow 90 100 100 82 78

PCyellow 100 83 85 78 80

greenR11 95 68 63 64 52

CCRIgreen 95 85 91 88 85

PCgreen 100 82 91 81 78

blueR12 92 56 53 61 72

CCRIblue 97 100 100 100 98

PCblue 100 93 89 86 81

Correlation among Ri, CCRIi and PC

Red

-20

0

20

40

60

80

100

LED75 LED51 LED40 LED28

Light sources

Color rendering index

R9

CCRIred

PCred

Correlation among Ri, CCRIi and PCYellow

0

20

40

60

80

100

120

LED75 LED51 LED40 LED28

Light sources

Color rendering index

R10

CCRIyellow

PCyellow

Correlation among Ri, CCRIi and PC

Green

0

20

40

60

80

100

LED75 LED51 LED40 LED28

Light sources

Color rendering index

R11

CCRIgreen

PCgreen

Correlation among Ri, CCRIi and PC

Blue

0

20

40

60

80

100

120

LED75 LED51 LED40 LED28

Light sources

Color rendering index

R12

CCRIblue

PCblue

Ra, CCRIa and percent correct

Light sources

D65 LED75 LED51 LED40 LED28

Tcp(K) 6420 7500 5100 4000 2800

Ra 96 84 79 80 75

CCRIa 95 95 95 91 87

PCa 100 80 82 77 74

Correlation among Ra, CCRIa and PC

Average

0

20

40

60

80

100

LED75 LED51 LED40 LED28

Light sources

Color rendering index

Ra

CCRIa

PCa

SUMMARIES

• Color rendering qualities of white LED (blue LED and yellow phosphor type) light sources of four different Tcp were examined with PC, Ri, and CCRIi.

• Correlation between CCRI and PC for different color temperature light sources is higher than those between CIE CRI and PC.

CONCLUSION

• If you like to take account of color categorization in order to evaluate practical color rendering quality of white LED light sources, the CCRI can be recommended rather than Ra.