Position des pions du testC 32 dans L*a*b*lorsqu'ils sont éclairés par les lum ières de l'expérience -20 -15 -10 -5 0 5 10 15 20 -30 -20 -10 0 10 20 30 a* b* Solux W W ARGB WR RGBA RGB Designing Designing a colour discrimination test to assess colour a colour discrimination test to assess colour rendering of rendering of LED sources. LED sources. Elodie Mahler – MST Optique Physiologique, Optique de contact et Optométrie, Paris 11, Orsay. Françoise Viénot – Centre de Recherche sur la Conservation des Documents Graphiques, MNHN, Paris. Jean-Jacques Ezrati – Centre de Recherche et de Restauration des Musées de France, Paris. Alain Bricoune – Lumières de Paris, Paris. Purpose Purpose Conclusion Conclusion Materials and Materials and Methods Methods Test: We manufactured 3 a desaturated Panel-like discrimination test with 32 caps equally distributed along a colour circle in CIELAB. L* = 78.54 ±0.69 C* ab = 15.05 ±2.16 E ab = 3.02±0.55 CIELAB units between adjacent caps. Figure 1. The colour discrimination test: the “Cercle 32”, designed to assess colour rendering of LED light sources. Light sources: - RGB LED cluster (RGB). - RGB+Amber LED cluster (RGBA). - two-phosphor cold white LED+Red LED cluster (WR). - two-phosphor cold white LED and two- phosphor warm white LED +Amber +RGB LED cluster (WWARGB). Relative intensities of each LED in every cluster were set in order to optimise the CIE Colour Rendering Indices (CRI) 4 . - Control light (Solux tungsten-halogen lamp). Table 1. Colour specification of the light emitted by the sources used in the experimentFigure 2. Spectral power distribution of the light emitted by the sources used in the experiment, measured using a Minolta CS-1000 spectroradiometer. Observers: 40 observers having normal colour vision as tested with Nagel anomaloscope. Each observer was tested with the five types of illumination sequentially. The sequence of illumination was balanced between observers. - Test duration : 3 minutes - Time interval between two illuminations : 3 minutes Results Results Discrimination test: - 7 / 40 observers successfully passed the test under all illuminations. - 33 / 40 observers failed under one or more illuminations. In order to quantify the impairment, we computed a discrimination Index based on the elevation of the average perceptible colour difference. The full colour path in L*a*b* was obtained by adding the 32 colour steps between two caps placed next to the other by the observer. The length of the correct colour path is of 96.78 CIELAB units. For example, the length of an erroneous colour path with an inversion between cap 17 & 18 is of 102.2. This makes an increase of the path of 5.6%. For RGB illumination, 46% of failures occur along the yellow axis, namely among greenish-blue shades and among purple shades, instead of 25% for the unbiased distribution. Figure 3. Examples of fails under each type of illumination, selected among observers plots. They are representative of the most frequent errors. Colour Rendering Index: We calculated the general and the special colour rendering indices according to the Test-Colour Method as recommended by CIE. In order to compare the CRI with the visual discrimination test results, we also computed a CRI based on the C32 test colour samples. Figure 4. Indices describing the discrimination efficiency and the CRI of the light sources. We performed statistical tests about the number of erroneous tests (χ²), and about the average path (), and found that, at the risk of 5%, the five light tests are not significantly different. RGB LED illumination yields poor discrimination and poor colour rendering. The most impaired colours by RGB LED light lie along the yellow axis direction, and are located in two areas of the hue circle, that is to say greenish-blue and purple shades. A supplement of light, as in the WWARGB LED cluster, should be offered in the missing parts of the spectrum to achieve the same success as the continuous spectrum light. The colour discrimination is correlated with the colour rendering. An improved CRI based on uniformly distributed test samples would make appear the deficiency axis and also provide colour discrimination indices. Discussion Discussion LED technology offers light sources with narrow spectrum in the visible range. Now, several LED arrangements allow the production of white light¹. To assess the quality of colour rendering of LED white light sources, we recently designed an experiment with real colour surfaces, light sources and observers. 57 colour normal observers performed the desatured Panel DD15 test from Lanthony illuminated with various LED clusters, and with a control continuous spectrum light². Compared with the control illumination, all LED clusters impaired colour discrimination, with a severe impairment for RGB LED cluster around greenish-blue and purple shades. However, the steps between adjacent caps of the DD15, that is designed to highlight confusion lines, are not constant in L*a*b*. The steps along the yellow axis are small. In order to avoid biased conclusions, we have designed a colour discrimination test aimed at evaluating lighting quality, to reveal low discrimination efficiency in any region of the hue circle. 1. CIE Expert Symposium on LED Light-sources, Physical Measurement and Visual and Photobiological Assessment, Tokyo, 2004. 2. Viénot F, Discriminating colours under LED illumination, AIC Grenade, 2005. 3. Leprêtre G., private manufacture of the tray, Paris 2005. 4. CIE, Method of Measuring and Specifying Colour Rendering Properties of Light Sources, Publication CIE 13.3, 1995. References References N um ber oferro neous tests /40 14 11 11 21 19 0 5 10 15 20 25 Solux W W ARGB WR RGBA RGB P ercentage o f co lo r p ath increase in L*a*b* 3,26 % 2,84 % 3,21 % 5,35 % 7,09 % 0 1 2 3 4 5 6 7 8 C olour R end ering Indices C IE R a8 R a14 6,32 95,50 96,78 96,63 93,98 88,77 91,35 67,36 60,34 21,71 0 10 20 30 40 50 60 70 80 90 100 C olour R endering Index C 32 caps 96,15 97,51 95,11 78,33 49,28 0 10 20 30 40 50 60 70 80 90 100 Solux W W ARGB WR RGBA RGB Solux W W ARGB WR RGBA RGB LED cluster LED cluster LED cluster LED cluster C hrom aticity coordinate x 0,385 0,382 0,379 0,385 0,381 C hrom aticity coordinate y 0,386 0,381 0,369 0,387 0,385 C olourtem perature T (k) 3960 4000 3990 4030 4030 C hrom aticity difference ΔC uv betw een the testsource and 0,0033 0,0017 0,0033 0,0042 0,0034 the reference illum inant Illum inance (lx) 653 669 657 657 666 Solux 0 0,002 0,004 0,006 0,008 0,01 0,012 380 480 580 680 780 W avelength [nm ] Spectral pow er distribution WR 0 0,004 0,008 0,012 0,016 0,02 380 480 580 680 780 W avelength [nm ] Spectral pow er distribution RGB 0 0,01 0,02 0,03 0,04 0,05 0,06 380 480 580 680 780 W avelength [nm ] Spectral pow er distribution RGBA 0 0,005 0,01 0,015 0,02 0,025 0,03 380 480 580 680 780 W avelength [nm ] Spectral pow er distribution W W ARGB 0 0,004 0,008 0,012 0,016 0,02 380 480 580 680 780 W avelength [nm ] Spectral pow er distribution C orrelation betw een the particularC 32 C R Is and apparentsteps in L*a*b*ofC 32 C aps under RGB 0 10 20 30 40 50 60 70 80 90 100 0 5 10 15 20 25 30 C 32 caps (colour circle) C R IC 32 norm alized /D elt C IE LA B units norm alized C R IC 32 Steps underR G B S olux O bs LD 1 2 3 4 6 5 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 25 24 26 27 28 29 30 31 32 a* b* error31 1 32 2 Errors ofneighbourhood ofeach cap 0 1 2 3 4 5 6 7 8 9 10 1 4 7 10 13 16 19 22 25 28 31 32 caps N um beroferrors ofneighbourhood Solux W W ARGB WR RGBA RGB Figure 5. Position of 32 caps colours in L*a*b* when illuminated by the sources used in the experiment and by the Planckian radiator at the same colour temperature. Figure 6. Number of erroneous choices of neighbouring cap, for each cap, including all examinations. The distribution of errors in the chromaticity circle is different between RGB LED illumination and Solux illumination. However, all illuminations have the same pattern of distribution approximately. Figure 7. Correlation between particular C32 CRIs and apparent steps between adjacent caps in L*a*b* of the C32 caps under RGB illumination The correlation between colour rendering and colour discriminating is fine. It is higher involving C32 or CIE CRI and average path increase than in any other combination of indices. Compared with the errors under control illumination, the errors under RGB LED light are often larger. Thus, rather than the number of erroneous tests, we look for an explanation of the path increase. Figure 5 shows that the chromaticity circle drawn for the C32 test is apparently widened for RGB LED illumination. Its average step between adjacent caps is of 3.86 ± 1.16. In opposition, it is of 2.84 ± 0.52 for Solux illumination. RGB illumination changes the specification of colours. It pushes to the extremity of an ellipse the greenish-blue and purple shades. Besides, the steps at the apices of the ellipse are smaller, although the average step increases. The discrimination efficiency of RGB illumination is reduced precisely for these falsely saturated colours (fig. 6). W W ARGB O bs O M -15 -10 -5 0 5 10 15 20 -20 -15 -10 -5 0 5 10 15 20 a* b* error20 22 21 23 WR O bs LD -15 -10 -5 0 5 10 15 20 -20 -15 -10 -5 0 5 10 15 20 a* b* error 28 30 29 31 RGBA O bs CC -15 -10 -5 0 5 10 15 20 -20 -10 0 10 20 a* b* error19 21 20 22 RGB O bs LL -15 -10 -5 0 5 10 15 20 -20 -15 -10 -5 0 5 10 15 20 a* b* error30 1 31 32 2 Position in L*a*b* of caps when illuminated with the light sources used in the experiment
Designing a colour discrimination test to assess colour rendering of LED sources. Elodie Mahler – MST Optique Physiologique, Optique de contact et Optométrie,
Jan 04, 2016
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Position des pions du test C32 dans L*a*b* lorsqu'ils sont éclairés par les lumières de l'expérience
-20
-15
-10
-5
0
5
10
15
20
-30 -20 -10 0 10 20 30
a*
b*
Solux WWARGB WR RGBA RGB
Designing Designing a colour discrimination test to assess colour rendering of a colour discrimination test to assess colour rendering of
LED sources.LED sources.Elodie Mahler – MST Optique Physiologique, Optique de contact et Optométrie, Paris 11, Orsay.
Françoise Viénot – Centre de Recherche sur la Conservation des Documents Graphiques, MNHN, Paris.
Jean-Jacques Ezrati – Centre de Recherche et de Restauration des Musées de France, Paris. Alain Bricoune – Lumières de Paris, Paris.
PurposePurpose ConclusionConclusion
Materials and Materials and MethodsMethods
Test: We manufactured3 a desaturated Panel-like discrimination test with 32 caps equally distributed along a colour circle in CIELAB.
L* = 78.54 ±0.69C*ab = 15.05 ±2.16Eab = 3.02±0.55 CIELAB units between adjacent caps.
Figure 1. The colour discrimination test: the “Cercle 32”, designed to assess colour rendering of LED light sources.
Light sources: - RGB LED cluster (RGB).- RGB+Amber LED cluster (RGBA).- two-phosphor cold white LED+Red LED cluster (WR).- two-phosphor cold white LED and two-phosphor warm white
LED +Amber +RGB LED cluster (WWARGB).Relative intensities of each LED in every cluster were set in
order to optimise the CIE Colour Rendering Indices (CRI)4. - Control light (Solux tungsten-halogen lamp).
Table 1. Colour specification of the light emitted by the sources used in the experiment.
Figure 2. Spectral power distribution of the light emitted by the sources used in the experiment, measured using a Minolta CS-1000 spectroradiometer.
Observers: 40 observers having normal colour vision as tested with Nagel anomaloscope. Each observer was tested with the five types of illumination sequentially. The sequence of illumination was balanced between observers.
- Test duration : 3 minutes- Time interval between two illuminations : 3 minutes
ResultsResultsDiscrimination test: - 7 / 40 observers successfully passed the test under all illuminations.- 33 / 40 observers failed under one or more illuminations.
In order to quantify the impairment, we computed a discrimination Index based on the elevation of the average perceptible colour difference. The full colour path in L*a*b* was obtained by adding the 32 colour steps between two caps placed next to the other by the observer. The length of the correct colour path is of 96.78 CIELAB units. For example, the length of an erroneous colour path with an inversion between cap 17 & 18 is of 102.2. This makes an increase of the path of 5.6%.
For RGB illumination, 46% of failures occur along the yellow axis, namely among greenish-blue shades and among purple shades, instead of 25% for the unbiased distribution.
Figure 3. Examples of fails under each type of illumination, selected among observers plots. They are representative of the most frequent errors.
Colour Rendering Index: We calculated the general and the special colour rendering indices according to the Test-Colour Method as recommended by CIE.
In order to compare the CRI with the visual discrimination test results, we also computed a CRI based on the C32 test colour samples.
Figure 4. Indices describing the discrimination efficiency and the CRI of the light sources.
We performed statistical tests about the number of erroneous tests (χ²), and about the average path (), and found that, at the risk of 5%, the five light tests are not significantly different.
RGB LED illumination yields poor discrimination and poor colour rendering.The most impaired colours by RGB LED light lie along the yellow axis direction, and are located in two areas of the hue circle, that is to say greenish-blue and purple shades. A supplement of light, as in the WWARGB LED cluster, should be offered in the missing parts of the spectrum to achieve the same success as the continuous spectrum light.
The colour discrimination is correlated with the colour rendering. An improved CRI based on uniformly distributed test samples would make appear the deficiency axis and also provide colour discrimination indices.
DiscussionDiscussion
LED technology offers light sources with narrow spectrum in the visible range. Now, several LED arrangements allow the production of white light¹.
To assess the quality of colour rendering of LED white light sources, we recently designed an experiment with real colour surfaces, light sources and observers. 57 colour normal observers performed the desatured Panel DD15 test from Lanthony illuminated with various LED clusters, and with a control continuous spectrum light². Compared with the control illumination, all LED clusters impaired colour discrimination, with a severe impairment for RGB LED cluster around greenish-blue and purple shades. However, the steps between adjacent caps of the DD15, that is designed to highlight confusion lines, are not constant in L*a*b*. The steps along the yellow axis are small. In order to avoid biased conclusions, we have designed a colour discrimination test aimed at evaluating lighting quality, to reveal low discrimination efficiency in any region of the hue circle.
1. CIE Expert Symposium on LED Light-sources, Physical Measurement and Visual and Photobiological Assessment, Tokyo, 2004.2. Viénot F, Discriminating colours under LED illumination, AIC Grenade, 2005.3. Leprêtre G., private manufacture of the tray, Paris 2005.4. CIE, Method of Measuring and Specifying Colour Rendering Properties of Light Sources, Publication CIE 13.3, 1995.
ReferencesReferences
Number of erroneous tests / 40
14
11 11
21
19
0
5
10
15
20
25
Solux
WWARGB
WR
RGBA
RGB
Percentage of color path increase in L*a*b*
3,26 %2,84 %
3,21 %
5,35 %
7,09 %
0
1
2
3
4
5
6
7
8
Colour Rendering Indices CIE Ra8 Ra14
6,32
95,5096,78 96,63 93,9888,7791,35
67,36
60,34
21,71
0
10
20
30
40
50
60
70
80
90
100
Colour Rendering Index C32 caps
96,1597,51 95,11
78,33
49,28
0
10
20
30
40
50
60
70
80
90
100Solux
WWARGB
WR
RGBA
RGB
Solux WWARGB WR RGBA RGBLED cluster LED cluster LED cluster LED cluster
Chromaticity coordinate x 0,385 0,382 0,379 0,385 0,381Chromaticity coordinate y 0,386 0,381 0,369 0,387 0,385Colour temperature T (k) 3960 4000 3990 4030 4030
Chromaticity difference ΔCuv between the test source and 0,0033 0,0017 0,0033 0,0042 0,0034
the reference illuminantIlluminance (lx) 653 669 657 657 666
Solux
0
0,002
0,004
0,006
0,008
0,01
0,012
380 480 580 680 780
Wavelength [nm]
Sp
ectr
al p
ow
er
dis
trib
uti
on
WR
0
0,004
0,008
0,012
0,016
0,02
380 480 580 680 780
Wavelength [nm]
Sp
ectr
al p
ow
er
dis
trib
uti
on
RGB
0
0,01
0,02
0,03
0,04
0,05
0,06
380 480 580 680 780
Wavelength [nm]
Sp
ectr
al p
ow
er
dis
trib
uti
on
RGBA
0
0,005
0,01
0,015
0,02
0,025
0,03
380 480 580 680 780Wavelength [nm]
Sp
ectr
al p
ow
er
dis
trib
uti
on
WWARGB
0
0,004
0,008
0,012
0,016
0,02
380 480 580 680 780
Wavelength [nm]
Sp
ectr
al p
ow
er
dis
trib
uti
on
Correlation between the particular C32 CRIs and apparent steps in L*a*b* of C32 Caps under RGB
0
10
20
30
40
50
60
70
80
90
100
0 5 10 15 20 25 30
C32 caps (colour circle)
CR
I C
32
no
rma
lize
d /
De
lta
C
IEL
AB
un
its
no
rma
lize
d
CRI C32 Steps under RGB
Solux Obs LD
12
3
4
6 5
789
1011121314
15
1617
1819
2021
2223 2524 26
272829
303132
a*
b*
error 31 1 32 2
Errors of neighbourhood of each cap
0
1
2
3
4
5
6
7
8
9
10
1 4 7 10 13 16 19 22 25 28 31
32 caps
Nu
mb
er
of
err
ors
of
ne
igh
bo
urh
oo
d
Solux WWARGB WR RGBA RGB
Figure 5. Position of 32 caps colours in L*a*b* when illuminated by the sources used in the experiment and by the Planckian radiator at the same colour temperature.
Figure 6. Number of erroneous choices of neighbouring cap, for each cap, including all examinations. The distribution of errors in the chromaticity circle is different between RGB LED illumination and Solux illumination. However, all illuminations have the same pattern of distribution approximately.
Figure 7. Correlation between particular C32 CRIs and apparent steps between adjacent caps in L*a*b* of the C32 caps under RGB illumination
The correlation between colour rendering and colour discriminating is fine. It is higher involving C32 or CIE CRI and average path increase than in any other combination of indices.
Compared with the errors under control illumination, the errors under RGB LED light are often larger. Thus, rather than the number of erroneous tests, we look for an explanation of the path increase.
Figure 5 shows that the chromaticity circle drawn for the C32 test is apparently widened for RGB LED illumination. Its average step between adjacent caps is of 3.86 ± 1.16. In opposition, it is of 2.84 ± 0.52 for Solux illumination. RGB illumination changes the specification of colours. It pushes to the extremity of an ellipse the greenish-blue and purple shades. Besides, the steps at the apices of the ellipse are smaller, although the average step increases.
The discrimination efficiency of RGB illumination is reduced precisely for these falsely saturated colours (fig. 6).
WWARGB Obs OM
-15
-10
-5
0
5
10
15
20
-20 -15 -10 -5 0 5 10 15 20
a*
b*
error 20 22 21 23
WR Obs LD
-15
-10
-5
0
5
10
15
20
-20 -15 -10 -5 0 5 10 15 20
a*b
*error 28 30 29 31
RGBA Obs CC
-15
-10
-5
0
5
10
15
20
-20 -10 0 10 20
a*
b*
error 19 21 20 22
RGB Obs LL
-15
-10
-5
0
5
10
15
20
-20 -15 -10 -5 0 5 10 15 20
a*
b*
error 30 1 31 32 2
Position in L*a*b* of caps when illuminated with the light sources used in
the experiment
Related Documents
