color tolerancingthe case for instrumental · 2001. 11. 6. · wyszecki robertson ganz macadam....

ROY S. BERNS

Munsell Color Science Laboratory Chester F. Carlson Center for Imaging Science

Rochester Institute of Technology54 Lomb Memorial Drive

Rochester, New York 14623-5604berns@cis.rit.edu

ROY S. BERNS

Munsell Color Science Laboratory Chester F. Carlson Center for Imaging Science

Rochester Institute of Technology54 Lomb Memorial Drive

Rochester, New York 14623-5604berns@cis.rit.edu

The Case for InstrumentalColor Tolerancing

The Case for InstrumentalColor Tolerancing

This talk was presented at a meeting of the Detroit Color Council on June 10, 1997 in Dearborn, Michigan. The theme for this meeting was color measurement and control of automotive materials. This presentation was supported by the Detroit Color Council.

This talk was presented at a meeting of the Detroit Color Council on June 10, 1997 in Dearborn, Michigan. The theme for this meeting was color measurement and control of automotive materials. This presentation was supported by the Detroit Color Council.

Outline of PresentationOutline of Presentation

• Introduction• Data Bases• Visualization of Systematic Trends• Model Development• Model Optimization• Conclusions

Two Competing Philosophies:Two Competing Philosophies:

Visual Instrumental

A Case for Instrumental QAA Case for Instrumental QA

PrerequisitesStatistical sampling of batch

Precise and accurate instrument

Presentation OutlineLimitations of CIELAB

Review of CMC and CIE94

Observer uncertainty

Recommendations

1973 Meeting of Colorimetry Committee1973 Meeting of Colorimetry Committee

City University, London

WyszeckiRobertson

Ganz

MacAdam

CIELABCIELAB"Pending the development of an improved coordinate system..." (CIE Pub. 15.2)

65432100

2

4

6

8

∆∆∆∆ E*ab

∆∆∆∆ V

875 color-difference pairs judged by 50 observers (43,750 judgements)

Systematic Limitations in CIELABSystematic Limitations in CIELAB

Color difference pairs with the same visual difference had ∆∆∆∆E*

ab values that depended on

the C*ab

of the standard

McDonald, 1974

∆E = ∆L2 + ∆C1+ 0.02C

2

+ ∆H2

1/2

C*ab

Sc

1.0

Sc

Society of Dyers and Colourists, 1984: CMCSociety of Dyers and Colourists, 1984: CMC

∆ECMC(l:c) = CF∆L*

lSL

2

+ ∆Cab*

cSC

2

+ ∆Hab*

SH

2

1/2

1008060402000.4

0.6

0.8

1.0

1.2

1.4

1.6

CIELAB L*

CM

C S

l

10050000

1

2

3

4

CIELAB C*

CM

C S

c

300200100001.0

1.2

1.4

1.6

1.8

2.0

CIELAB hue angleC

MC

Sh

CIELAB C* = 50

Wound thread

CIE Division 1 Meeting: Buenos Aires, 1989CIE Division 1 Meeting: Buenos Aires, 1989

TC 1-27 Color differences - soft copy & hard copy

TC 1-28 Parametric factors

TC 1-29 Industrial color difference equation

TC 1-27 Color differences - soft copy & hard copy

TC 1-28 Parametric factors

TC 1-29 Industrial color difference equation

TC 1-29 PhilosophyTC 1-29 PhilosophyContinuous improvementApplicable to all industriesStatistical validityTimely results

k = parametric factor / functions = corrective function

∆ETC1−29* = ∆L*

kLSL

2

+ ∆Cab*

kCSC

2

+ ∆Hab*

kHSH

2

1/2

Statistically Valid DataStatistically Valid Data

Requirements: Multiple observers, precision estimates, surface colors

• WittPainted samples, hairline separation, 20%Y background, 22-24 observers

• Luo-RiggWool serge, hairline separation, 14.2%Y background, 20 observers

Goal to combine 13 experiments each of about 20 observers

• RIT-DupontPainted samples, hairline separation, 10.9% background, 50 observers

Visualization of Systematic Trends: Lightness R 2=0.05Visualization of Systematic Trends: Lightness R 2=0.05

1008060402000

1

2

3

Luo-Rigg

RIT-Dupont

Witt

L*

Del

ta L

*

• Luo-Rigg: Random

• Witt and RIT-Dupont:

Consistent and different

1008060402000

1

2

3

L*

Fun

ctio

n va

lue

CMC

BFD

RIT-Dupont

Optimized Functions(not normalized)

Visualization of Systematic Trends: Chroma R 2=0.70Visualization of Systematic Trends: Chroma R 2=0.70

1201008060402000

1

2

3

Luo-Rigg

RIT-Dupont

Witt

C*

Del

ta C

*

• Consistent trends

for all data sets

• Main reason

for improvement from:

Mc Donald, 1974

Mc Laren, 1976

JPC79, 1979

CMC, 1984

BFD, 1987

Visualization of Systematic Trends: Hue R 2=0.45Visualization of Systematic Trends: Hue R 2=0.45

1201008060402000.0

0.2

0.4

0.6

0.8

1.0

Luo-RiggRIT-Dupont

Witt

C*

Del

ta H

* • Upward Trend

• Weaker Than Chroma Trend

Visualization of Systematic Trends: Hue Angle R 2=0.01Visualization of Systematic Trends: Hue Angle R 2=0.01

40030020010000.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Luo-RiggRIT-Dupont

Witt

h

Del

ta

Hue

/(1+

0.01

5C*s

td)

Corrected for C* position

• Random Dispersion

• No Consistent Trend

TC 1-29 ConclusionsTC 1-29 Conclusions

• SL other than unity would fit parametric

factors• S

L other than unity would not be universal

across industries• S

C and S

H functions should be linear due

to data dispersion• S

H function should not have a hue angle

dependency

CIE94 EquationCIE94 Equation

CIE 116-1995

∆E94* =

∆L*

kLSL

2

+ ∆Cab*

kCSC

2

+ ∆Hab*

kHSH

2

1/2

SL = 1

SC = 1 + 0.045Cab*

SH = 1 + 0.015Cab*

1008060402000.0

1.0

CIELAB L*

Fu

nct

ion

va

lue

CMC

CIE94

-15 -5 5 15-5 5 15

-15

-5

5

15

∆a*∆

L*

CMC vs. CIE94 Lightness ToleranceCMC vs. CIE94 Lightness Tolerance

CMC

Who's right?Who's right?

Optimizing TolerancesOptimizing Tolerances

22 observers making "pass/fail" decisions for 32 samples distributed about a standard

- 1 5 - 5 5 1 5- 1 5

- 5

5

1 5

∆ a *

- 1 5 - 5 5 1 5

- 5

5

- 1 5 - 5 5 1 5- 1 5

- 1 5 - 5 5 1 5∆ a *

- 1 5 - 5 5 1 5

1 5

Optimized Ellipsoids for Four Observers, All Rejecting 18 Samples!

Optimized Ellipsoids for Four Observers, All Rejecting 18 Samples!

Who's right?Who's right?

Number of samples that were unacceptable

Num

ber

of o

bser

vers

0

1

2

3

4

13

14

15

16

17

18

19

20

21

22

23

24

25

Optimized Ellipsoids for Three ObserversOptimized Ellipsoids for Three Observers13, 19, and 25 samples rejected

∆b*

∆a*

∆L*

∆a*

∆L*

∆b*

Who's right?Who's right?

Where Are We?Where Are We?

CMC CIE94Optimize visual/instrumental data"Tomorrow is another day"

We Understand the Primary Limitation of CIELABWe Understand the Primary Limitation of CIELAB

Instrumental Color Tolerancing Can be Accomplished

1201008060402000

1

2

3

Luo-Rigg

RIT-Dupont

Witt

C*

Del

ta C

*

1008060402000

1

2

3

4

5

6

C *Fu

ncti

on v

alu

e CIE94

CMC

ab

Choose an Appropriate EquationChoose an Appropriate Equation

Be Conservative!

CIE94 unless you KNOW betterCIE94 unless you KNOW better

Only Adjust the Commercial FactorOnly Adjust the Commercial Factor

One observer is never representative of a population.

- 1 5 - 5 5 1 5

- 5

5

- 1 5 - 5 5 1 5- 1 5

- 1 5 - 5 5 1 5∆ a *

- 1 5 - 5 5 1 5

1 5

Finally,Finally,

empower yourself through understanding

Thanks DCC for sponsoring this meeting.

Thanks DCC for sponsoring this meeting.

And...And...

Support research that leads to improved equation development.

Munsell Color Science Laboratory Industrial Color Difference

Consortium

Munsell Color Science Laboratory Industrial Color Difference

Consortium

For example, join the:

Current Members:Current Members:

3MBayerDatacolorDetroit Color CouncilDupontDystarGretag/MacbethInter-Society Color CouncilPPGSociety of Plastics EngineersXerox