Color Quality Scale (CQS): Measuring the color quality of light sources Wendy Davis wendy.davis@nist.gov O ti l T h l Di i i Optical Technology Division National Institute of Standards and Technology
Copyright Materials This work was prepared by an officer or employee of the United States t government as part of that t person's official i duties. It is not subject to copyright in the United States and there are no copyright restrictions on reproduction, derivative works, distribution, performance, or display of the work. Anyone may, without restriction under U.S. copyright laws: F i f ti For more information see: http://www.usa.gov/copyright.shtml
Learning Objectives History of the Color Rendering Index (CRI) what CRI actually measures how CRI is calculated Problems & limitations of the CRI why there is strong motivation to replace it Dimensions of color quality various approaches to evaluating color quality Color Quality Scale (CQS) international standardization work
Color Rendering Equal Energy Blue & Yellow LED
Importance of Measuring Color Rendering Daylight White LED Low pressure sodium lamp Theoretical maximum ~ 250 lm/w ~ 400 lm/w 520 lm/w Excellent?? No color color rendering color rendering rendering
History: CIE Progress 1948: 8 band Spectral Band Method (SBM), deviation from full radiator 1955: established WC 1.3.2 to address terminology and compare SBM with test sample method 1961: agreed on test t color method, with 8 test t samples 1964: Publication 13 (1st edition), published test sample method
History: CIE Progress 1974: Publication 13 (2nd edition) Defined reference illuminants Test samples: 8 + 6 Von Kries chromatic adaptation transform Use the CIE 1964 UCS Scaling: Warm white halophosphate lamp to Ra=50 Developed e the CRI as we know it today
Color Rendering Index (CRI) Test source Reference source Same CCT [K] CIE Dxx 400 500 600 700 Planckian (CCT<5000 K) Standard Daylight (CCT > 5000 K) 400 500 600 700 #1 #2 #3 #4 #5 #6 #7 #8 0.8 0.7 0.6 05 0.5 TCS01 TCS03 TCS05 TCS07 TCS02 TCS04 TCS06 TCS08 R a 0.3 0.2 0.1 0.4 0 350 450 550 650 750 #9 #10 #11 #12 #13 #14 TCS09 TCS10 TCS11 TCS12 0.9 TCS13 TCS14 0.8 TCS15 0.7 0.6 0.5 0.4 0.3 0.2 01 0.1 0 350 450 550 650 750
Aside: Measuring Color Differences Two-dimensional diagrams Only for light color No black, grey, or brown Three attributes of color are hue, chroma (saturation), and lightness, and are expressed in a three dimensional space. Chroma Lightness white To allow accurate specification of object colors and color differences, CIE recommended CIELAB and CIELUV in 1976. black Hue
Color Rendering Index (CRI) Keep perspective: The CRI was designed to evaluate fluorescent lamps The CRI was intended to measure the naturalness of objects colors Reflective samples (1 st eight) chosen to represent average saturation of objects Has had no substantive changes in 35 years
Problems with the CRI High scores do not guarantee good saturated colors RGB model (Ra=80) poor color - 466/538/603 3-LED Model Peaks at: 457, 540, & 605 nm 400 500 600 700 CRI =80 Ref. LED Products are optimized for metrics. Inadequate metrics can lead to bad products.
Problems with the CRI CRI penalizes light sources having enhanced color contrast Neodymium incandescent lamp CRI = 77 (Normal incandescent lamp CRI=100) Ref. LED
Problems with the CRI RGB white LEDs can have the same effects 3-LED Model Peaks at: 464, 538, 620 nm CRI = 63 Ref LED Products are optimized for metrics. Products are optimized for metrics. Outdated metrics can impede development of new technologies.
Saturation & Naturalness Hunt Effect Colorfulness / saturation increases with luminance If we want objects to appear most natural in perceived color, we may want artificial sources to enhance their saturation.
Color Saturation Color Preference Deane Judd Flattery Deane Judd, Flattery Index for Artificial Illumination (1967)
Approaches to Measuring Preference Other metrics Use method very similar to CRI, but penalties for color differences are based on deviations from preferred shifts. Flattery Index (Judd, 1967) Color Preference Index (Thornton, 1974) Examples of preferred shifts: Skin tones (Caucasian) Foliage (green) more reddish less yellowish
Other Color Quality Approaches Two main presumptions: Gamut-based metrics Increased gamut increases discriminability between colors Increased gamut leads to increased object chroma Color Discrimination Index (Thornton, 1972) Cone Surface Area (Fotios, 1997) Color Rendering Capacity (Xu, 1993) Feeling of Contrast Index (Hashimoto, Yano, & Nayatani, 2000)
History: CIE Progress 1980s: new Technical Committee worked on subject, but closed without recommendations due to member disagreements 1995: Publication 13 (3 rd edition), no substantive changes from 2 nd edition, just fixed some errors 1999: Another Technical Committee (TC 1-33) closed without reaching consensus. Industry opposed proposed new procedures due to lack of visual experiments
To replace the CRI Color Quality Scale (CQS) Fix the problems of the CRI Replace outdated formulae in CRI Works for all light sources Considers not only color fidelity but also color preference and other aspects of color quality Initially developed with colorimetric simulations Being tested with vision science experiments
Color Quality Scale (CQS) Inspiration taken from CRI Uniform object color space to calculate color differences Test sample method Reference source, matched in CCT Single number output Primary deviations from CRI Name Different reflective samples Updated object color space & chromatic adaptation transform Saturation ti factor CCT factor RMS combining of color differences 0-100 scale New scaling factor
Reflective Sample Set CIELAB b* 80 60 New set of 15 20 saturated color 0 samples -20 40-80 -60-40 -20 0 20 40 60 80-40 -60-80 a* 15 new samples CIE 13.3 3 8 samples
Root Mean Square The CRI makes it possible for a lamp to score quite well, even when it renders one or two samples very poorly. This situation is more likely with SPDs having narrowband peaks, such as LEDs. RMS (Root Mean Square) color difference of 15 samples: ΔE RMS = 1 15 15 i=1 ΔE i 2 Example Case A Case B ΔE ΔE TS1 3 1 TS2 3 1 TS3 3 1 TS4 3 1 TS5 3 1 TS6 3 1 TS7 3 1 TS8 3 1 TS9 3 1 TS10 3 1 TS11 3 1 TS12 3 1 TS13 3 1 TS14 3 16 TS15 3 16 Ra(mean) 91 91 Ra(RMS) 91 82
Saturation Factor Ref Test Ref Test Score is decreased for the full color difference. Score is not penalized for increase of chroma. (Score is decreased for hue & lightness shifts)
CCT Factor The reference source at extremely low or high CCTs should not have perfect color rendering (score=100). b* CIELAB 80 60 40 20 0-80 -60-40 -20 0 20 40 60 80-20 -40-60 -80 a* 1500 K 2000 K 3000 K 6500 K CCT (K) Gamut Area Multiplication Factor 1000 1579 0.19 1500 5293 0.64 2000 7148 0.87 2500 7856 0.96 2856 8085 0.98 3000 8144 0.99 3500 8267 1.00 4000 8322 1.00 5000 8354 100 1.00 6000 8220 1.00 6500 8210 1.00 7000 8202 1.00 8000 8191 1.00 9000 8185 1.00 10000 8181 1.00 15000 8180 1.00 20000 8183 1.00
Some Results of the CQS Cool White Neodymium RGB (enhanced) Fluorescent Incandescent (464/538/620 nm) RGB (poor red) (457/540/605 nm) RGB model (Ra=80) poor color - 466/538/603 400 500 600 700 CRI =63 CQS=64 CRI =77 CQS=88 CRI =63 CQS=80 CRI =80 CQS=74 Consistency of scores is maintained for fluorescent lamps
Spectrally Tunable Lighting Facility (STLF)
STLF Specifications 22 color channels using 1800 high power LEDs/unit (to be expanded) Each color channel: 5 to 10 W optical power. Full power total optical power ~200 W, luminous flux ~50,000 lm. 500 lx illumination for most simulated spectra. Built by contract with Philips Color Kinetics. Built with NIST Director s funding.
Small Experiment 8 Subjects: 4 Caucasians, 3 Asians, 1 Hispanic 5 females, 3 males Tasks: 1) Judge color appearance of objects in the room 2) Judge color appearance of skin (hands and face)
Experimental Spectra Lights tested CCT: 4000 K (800 lx), Duv=0.000 3000 K (300 lx), Duv=0.000 4000 K 3000 K Reference: broadband d Ra=98 Ra=97 4 peak (RGBA) Desaturating-4 Ra=70 Ra=70 4 peak (RGBA) Desaturating-3 Ra=77 Ra=76 4 peak (RGBA) Desaturating-2 Ra=84 Ra=84 4 peak (RGBA) Desaturating-1 Ra=90 Ra=90 4 peak (RGBA) Neutral Ra=92 Ra=95 4 peak (RGBA) Saturating-1 Ra=87 Ra=88 4 peak (RGBA) Saturating-2 Ra=82 Ra=81 4 peak (RGBA) Saturating-3 Ra=74 Ra=74 4 peak (RGBA) Saturating-4 Ra=69 Ra=69
4000 K Broadband Reference
4000 K Desaturating-4
4000 K Desaturating-3
4000 K Desaturating-2
4000 K Desaturating-1
4000 K Neutral
4000 K Saturating-1
4000 K Saturating-2
4000 K Saturating-3
4000 K Saturating-4
Results: Room Evaluation
Results: Skin Evaluation
Results: Correlation Coefficients CRI Ra CQS Qa Room 4000 K 0.47 0.88 Room 3000 K 028 0.28 072 0.72 Skin tone 4000 K 0.21 0.73 Skin tone 3000 K 043 0.43 080 0.80 Room (both CCTs) 0.37 0.79 Skin tone (both CCTs) 0.31 0.75
Current Standardization Work Recent CIE Activity 2007: TC 1-62 published CIE 177 Studied color rendering of white LED sources Recommended the development of a new color rendering metric (or set of metrics) Recommended concurrent use of CRI and new metric, at least at first The new metric should be applicable to all types of flight sources
Current Standardization Work TC 1-69: Colour Rendition by White Light Sources Started in late 2006 Chair: Wendy Davis, US Terms of Reference: To investigate new methods for assessing the colour rendition properties of white-light sources used for illumination, including solid-state light sources, with the goal of recommending new assessment procedures.
Membership: TC 1-69 38 active members from 13 countries + a handful of interested observers Meetings: Physical meetings held annually Electronic communications critical: web site for sharing papers, data, etc. e-mail list with accessible archive web meeting in 2010
TC Research 17 (+/-) research reports 10 labs/groups seven countries Topics included: rendering of human skin chromatic discrimination color memory for real objects attractiveness & naturalness of fruits & vegetables estimations of color differences color harmony more.
Current Standardization Work Seven metric proposals: Rank-order based color rendering index (RCRI) Feel of contrast index (FCI) CRI-CAM02UCS Color quality scale (CQS) Harmony rendering index (HRI) Categorical color rendering index (CCRI) Memory CRI (MCRI)
Metric Proposals Rank-order based color rendering index (RCRI) Bodrogi, Bruckner, & Khanh Premise: An index with discrete increments are easier for users to understand 17 test color samples CAM02-UCS Feel of contrast index (FCI) Hashimoto, Yano, & Nayatani Premise: High chroma of colored objects = high visual clarity Derived by a simple transformation of gamut area Intended to supplement CRI
CRI-CAM02UCS Li, Luo, & Li Metric Proposals Premise: CRI-like method with updated colorimetry (UCS, CAT) Same test color samples as CRI CAM02-UCS Color quality scale (CQS) Davis & Ohno Premise: One metric can integrate different dimensions of color quality Does not treat all object color shifts equally 15 test color samples
Metric Proposals Harmony rendering index (HRI) Szabo, Bodrogi, & Schanda Premise: To quantify the distortion of color harmony caused by light sources Studied harmony with two- and three-color combinations under different illumination conditions CAM02-UCS Categorical color rendering index (CCRI) Yaguchi, Endoh, Moriyama, & Shioiri i i Premise: Errors in color rendering will shift the categorical color names of some object colors
Metric Proposals Memory CRI (MCRI) Smet, Forment, Hertog, Deconinck, & Hanselaer Premise: Memory colors of familiar objects used as reference to evaluate color rendering Based on experiments with real objects Ratings of similarity between apparent object color to memory color of the object
Current Standardization Work Final decision making/voting should happen very soon. Ideally, it would have already happened. Honestly, building consensus is proving challenging.
Thank your for your time! QUESTIONS?? This concludes The American Institute of Architects Continuing Education Systems Program Wendy Davis: wendy.davis@nist.govdavis@nist