How Modern Displays Push Conventional Colorimetry to Its Limit Abhijit Sarkar*, **, Laurent Blondé*, Patrick Le Callet**, Florent Autrusseau**, Patrick.

How Modern Displays Push Conventional Colorimetry to Its Limit

Abhijit Sarkar*,**, Laurent Blondé*, Patrick Le Callet**,

Florent Autrusseau**, Patrick Morvan*, Jürgen Stauder*

CREATE 2010, Gjøvik, Norway

June 8, 2010

*Technicolor Research & Innovation , France

**IRRCCyN-IVC, University of Nantes, France

1

About Me…

2

B.E. Electrical Engineering, Jadavpur University,

India, ‘00

M.S. Architectural Engineering (Lighting), Penn

State University, Pennsylvania, ‘05

MS Color Science, Munsell Color Science Lab,

RIT, Rochester, ’08

Currently: PhD Student at

• Technicolor Research & Innovation, Rennes, France

(Advisors: Laurent Blondé, Jürgen Stauder)

• University of Nantes , France (Advisors: Patrick Le

Callet & Florent Autrusseau)

Current Research Topic:

Observer variability issue in modern display

colorimetry

What is Color? How Do We Measure It?

Color is a perception that depends on the response of the human

visual system to light, and the interaction of light with objects

(also called color stimulus)

3

0.15470.70290.1424

Spectral Power Distribution

Spectral Sensitivity of

Photoreceptors

Tristimulus Values

Quantitative Representation of Color

ySk

zRSkZ

yRSkY

xRSkX

100

: Spectral power distribution of a CIE illuminant (e.g. D65) S

R

zyx ,,

: Spectral reflectance factor of the object

: CIE standard observer 2° or 10° color-

matching functions

ZYX

Xx

Common representation: xyY

4

Similar for y and z

In colorimetry, there is a provision for changing the illuminant and the object, but not the observer!

5

Average Observer

Observer Metamerism: The color is in your eyes!

Two color stimuli with very different spectral power

distributions can be a match for one observer and mismatch

for the other

6

The more different the spectral power distributions are, the more prominent the effect of metamerism is!

7

Observer Metamerism on a CRT and an LCD: Simulation

XYZ (10° SO and

D65)

dcCRT/LCD(10°)

XYZCRT/LCD(obs1)

M-1CRT/LCD (10°)

XYZCRT/LCD(obs2)

dcCRT/LCD_obs1

(CRT equiv.)

dcCRT/LCD_obs2

(CRT equiv.)

MCRT/LCD (obs2)MCRT/LCD (obs1)

M-1CRT (obs1) M-1

CRT (obs2)

CRT LCD

8

Obs. 1 (cat. 5)

Obs. 2 (cat. 6)

9

CRT LCD

Obs. 1 (cat. 5)

Obs. 2 (cat. 6)

A Color Matching Experiment Using Two Displays

10

Displays: a narrow-band LCD with LED backlight and a

broad-band studio CRT

Observers asked to adjust the color on the left half of

the bipartite field (matching field) to match the color on

the right half (test field)

text

LCD

CRT

10° Mask

Mirror

10°

Projector

Mask

Preliminary Results from the Color Matching Experiment

Standard observer-predicted color differences (∆E*00) of

individual observer color matches found to be significant

for some observers 11

Conclusions

12

In highly color-critical applications using modern displays,

employing an average observer model in colorimetry can

possibly lead to unacceptable color matches for many

color-normal expert observers

Objective of this work:

To develop an observer-dependent color imaging method,

where color workflow in a display device can potentially be

tuned to one of several observer classes, leading to novel

industrial applications

Thank you!

13

Acknowledgments

Jean-Jacques Sacré for his help and guidance in preparing the

experimental setup

Following colleagues for being the observers in the color matching

experiments:

• Laurent Blondé• Patrick Morvan• Jürgen Stauder• Arno Schubert• Pascal Bourdon• Séverine Baudry• Emmanuel Jolly• Catherine Serre• Jean-Jacques Sacré

Hypotheses: Solution Lies in the Problem Itself…

14

Color vision of observers can be classified into a small

number of categories

A practical experimental setup can be built to classify

observers

Color processing in an industrial application can be tuned

to one of these categories, leading to a better agreement

between real observer perceptions and colorimetry

Work-in-Progress: A Novel Experimental Method for Observer Classification

15

Seven Observer Categories (sets of CMFs) + CIE 10° observer

Rate categories for many colors:unacceptable, acceptable and

satisfactory

Determine observer category

Observer Categories Through Statistical Analysis of Color-Matching Dataset

16

4 x- functions 3 y-

functions

3 z- functions

Results: Observer Classification

17

For eleven observers, the CIE 10° standard observer (category

1) was not among the two most preferred categories

For three observers, category 1 was rejected as an

unacceptable match for all fifteen test colors

Conclusions

18

In highly color-critical applications using modern displays,

employing an average observer model in colorimetry can

possibly lead to unacceptable color matches for many color-

normal expert observers

Real, color-normal observers can be classified into a small

number of categories by means of a practical experimental

setup suitable for industrial applications

Objective of this work:

To develop an observer-dependent color imaging method, where

color workflow in a display device can potentially be tuned to one

of several observer classes, leading to novel industrial applications

Thank you!

19

Acknowledgments

Jean-Jacques Sacré for his help and guidance in preparing the

experimental setup

Following colleagues for being the observers in the color matching

experiments:

• Laurent Blondé• Patrick Morvan• Jürgen Stauder• Arno Schubert• Pascal Bourdon• Séverine Baudry• Emmanuel Jolly• Catherine Serre• Jean-Jacques Sacré

Appendix: Modern Displays and Observer Metamerism

20

CRT LCD

Obs. 1 (cat. 5)

Obs. 2 (cat. 6)

X Y Z X Y Z

Obs 1 35.11 39.18 44.50 36.71 40.82 46.50

Obs 2 36.07 39.18 43.48 38.66 40.84 45.29

X Y Z X Y Z

Obs 1 53.94 53.45 52.57 55.91 55.53 54.31

Obs 2 56.84 53.45 51.33 60.19 55.57 52.88

CRT LCD

All colors refer to CRT

Obs. 1 (cat. 5)

Obs. 2 (cat. 6)