You are on page 1of 10

NTC Project: C04 NS11

1
Optimizing Color Control Throughout the Supply Chain

Project Team:
Leader: David Hinks, North Carolina State University
Members: Dick Aspland, Clemson University
Renzo Shamey, North Carolina State University
Nancy Cassill, North Carolina State University
Warren Jasper, North Carolina State University
Rolf Kuehni, North Carolina State University
Project Website: http://www.ntcresearch.org/projectapp/index.cfm?project=C04-NS11

Goal
The primary goal of the proposed research is to develop an accurate and precise integrated color control
system that can be easily implemented throughout the US textile industrial manufacturing complex, from
product designer through to merchandiser, dyer, retailer and consumer.

Abstract
To date, this project has generated one MS thesis in Textile Chemistry [1], one peer-reviewed paper, 12
conference and symposium papers. Additionally, four peer-review manuscripts will be submitted in
October, 2007 [2-18]. We anticipate that at least three more peer-reviewed papers and a patent disclosure
will be submitted in 2008, and at least two international conference papers. The project also received the
1
st
place award for best poster in the chemistry division at the 2006 NTC Forum. While this project
formally ended in April 2007, research will continue into 2008 with the on-going Ph.D. research of Lina
Crdenas. Our website will be updated as papers are published. To achieve the stated goals we are
conducting experiments on several levels (expanded from our last NTC report) including:
Determination of the minimum variability in visual assessment of small color differences in textile
samples via the first international replication study.
Determination of the performance of key color difference models in the blue region of color space.
Assessment of key variables in the textile supply chain and the generation of a fishbone diagram.
Assessment of lighting in standard viewing booths and retail stores.
Assessment of variability in unique selection.
Assessment of the impact of human lens yellowing on color difference perception.

This report highlights some of the findings to date, while the cited papers report the data in detail. The
following work has been completed in the last 12 months.
1. Development of a perceptually linear scale, with corresponding visual assessments.
2. Assessment of the effects of gray scale size, and positioning (e.g., a gap between test samples and
gray scale reference pairs).
3. Assessment of the differences between nave and expert visual assessors for color difference.
4. Complete assessment of color difference performance for a comprehensive new visual dataset
around one blue color center.
5. Assessment of the scope and limitations of a custom composite statistical model, called
performance factor (PF/3).
6. Assessment of the variability in assessments of unique hues, which are a critical part of color
spaces and color appearance models.
7. Assessment of the effect of aging of the human lens on color difference perception.

On-going work is now focused on conducting an international replication experiment based on the
optimum visual assessment method developed through the project to date. Funding permitting,
participating countries in the replication experiment are anticipated to be: US, UK, Czech Republic, and
Hong Kong. The replication experiment will provide the most comprehensive replicated visual assessment
data for small color difference. The next stage (requiring further funding) will be to obtain highly
controlled visual data throughout the key regions of color space and develop either a new color space or
new formula for the assessment of color difference. This will be a significant but critical undertaking
fundamental to optimizing digital communication and management of color in the textile supply chain.
National Textile Center Annual Report: November 2007
NTC Project: C04 NS11
2
Introduction
Color is an essential part of the textile industry. It greatly influences sales volume and is a primary factor
in the purchase of a product. A common objective in color technology is to control and reproduce a color
under a set of specified conditions. In the textile industry, effective color control and communication
between designer, dyer and retailer are critical to obtaining high product quality and cost efficiency.
However, many of the factors enabling color control within the textile supply chain are not optimum, and
therefore it is likely that the critical path from product concept to consumer is extended, with associated
cost inefficiencies. This project attempts to address the major issues that will lead to optimization of color
control in the U.S. textile supply chain.

Visual and Colorimetric Assessment of Color Difference
An objective color difference formula accurately representing average perceptual assessments of observers
is a desirable tool for color quality control of textile materials, and is arguably critical for effective
electronic communication of colorimetric data for color management of a product supply chain. Existing
formulae are based on several different sets of perceptual data that have been established under various
experimental conditions, using samples representing a diverse range of substrates and different groups of
observers. In the textile industry the CMC (2:1) color difference formula is used as a standard metric [19-
21]. Recently, however, the International Commission on Illumination (CIE) recommended the
CIEDE2000 formula [12-13]. Luo et al. reported accuracy of prediction for several formulae against
average data from a large visual data set that combined four separate experimental data sets. The large data
set was used in the development of the CIEDE2000 formula. Using the PF/3 performance method, a value
of 67.4 for the CIEDE2000 formula vs. 62.1 for CMC (1:1) was reported [13].

While the new formula produced an improvement the results remain unsatisfactory. Four subsequent
independent field tests of CIEDE2000 vs. CMC (2:1) based on textile samples resulted in a similar level of
accuracy for the two formulae [14-16]. No data can be found in the literature that provides a definitive
answer to the disparity between the theoretical performance and the field-tested performance. However, our
hypothesis is that the differences lie primarily in the variability among visual assessment protocols used for
establishment and testing of the various color difference formulae developed.

There are many variables that affect the degree of accuracy in assessing color difference formulae.. This
part of our project was focused on identifying and minimizing the variables in visual assessment of small
color difference of textile materials and establishing the optimum level of intra- and inter-observer
variability. These data will be used to determine the maximum performance of any color difference model.
Once the best experimental conditions (for textile samples) are established, highly controlled replication
experiments will be performed under identical conditions in different regions of the world (US, Europe and
Asia).

One of the specific issues addressed in this report is: Do nave and expert observers differ in terms of intra-
and inter-observer variability in perceptual color difference assessments? (A nave observer in this case is
defined as an observer with no prior knowledge of commercial pass/fail color difference assessment,
experts are defined as color normal observers whose employment involve, or has involved, commercial
shade matching in the textile industry). Two further issues that we report are ongoing work toward the
development of unique gray scales to assess which visual method provides the least variance in data, and an
assessment of statistical methods for assessing agreement between inter and intra-observer datasets, and
observer-colorimetric models.

Survey of Color Issues in the Retail Industry
As stated in previous reports, we are continuing to survey the general needs and direction of the major US
and some foreign retailers in regard to product development. In all cases, it is clear that color control
constitutes the most difficult aspect of the supply chain. Some of the larger US retail chain stores (store
names will not be published due to sensitivity of the information) have stated that the product development
cycle is approximately 45 weeks, from concept-to-consumer. However, recently we reviewed the
manufacturing process of a leading UK retailer, and they are achieving product cycles of small production
lots of less than 11 weeks. This means that color control must be completed entirely via digital
communication, and requires significant technical expertise in colorimetry at each stage of the supply
National Textile Center Annual Report: November 2007
NTC Project: C04 NS11
3
chain. The UK retailer informed us that they necessarily accept that some of the products will have less
than desirable color differences. However, the ability to rapidly test markets and respond within a few
weeks makes the retailer highly dynamic and has been shown to increase consumer satisfaction and,
consequently, increases the number of store visits. One goal is to have new products in each store every
month. We have not yet found a US retailer with the capability for very rapid concept-to-consumer product
cycles.

In order to achieve these goals, the upstream processes, such as design, color standard development,
approval of first lab and production dyeings, and garment manufacture, will come under increased pressure
to deliver first quality product within a few weeks rather than months. We believe that significant advances
in color control - including visual assessment, color measurement, color difference calculations, and
lighting control are essential. In addition, it is clear that short product cycles provide a major opportunity
to all wet and dry processing that is geographically close to the US retail distribution centers (i.e., dyeing
and finishing within the US itself), and which has significant technical expertise. For example, it currently
takes approximately 10-12 weeks to ship textile products from mills in China to the Europe and the US, not
including processing time.

Therefore, US textile processing should embrace reducing the time frame for concept-to-consumer; digital
communication is a new paradigm that will enhance the competitiveness of capable sectors of the US
textile industry. It is clear that utilizing digital color communication (i.e., no sharing of physical swatches,
only sharing digital color data for pass/fail decisions) throughout the various stages of the supply chain is
an attractive route to shortening lead times and reducing costs.

However, in order for digital communication to be effective, the current mathematical models used to
correlate to visual assessment of color difference of textile materials must be improved substantially. It is
clear that models that predict absolute color, color difference, and color changes as a result of changes in
lighting (color inconstancy) are not optimum. The primary goal of the proposed research, therefore, is to
develop the fundamental data that quantifies the current variability in color control in the textile industry
and then to optimize color models and methods that will enable effective digital color communication
throughout the supply chain.

One of the key areas of this project is development of the first statistically valid international visual
replication study on textiles. This experiment will determine the minimum variability within a dataset, and
between observer datasets, for practical visual assessment of textile materials, which will lead to
quantification of the maximum performance of an optimized color difference formula. The need for the
replication visual study arises from new strategic requirements by the retail industry for very rapid product
cycles.

Key variables in textile color control
A practical approach to understanding the complexity of the number of variables in color control involves
the use of a fishbone flow chart, otherwise known as a cause and effect diagram. The fishbone diagram is a
tool to identify the potential or real causes that contribute to a single outcome. The causes are organized in
order of significance creating a hierarchical organization in relation to the outcome. In the case of color
communication, a huge array of factors exist that need to be clearly identified and understood. Ultimately,
work should focus on minimizing the variability of each factor. A fishbone diagram was used to depict
some of the most important factors in the communication and control of color in the textile supply chain.
The variability in the control of color through the supply chain was broken down into the following
categories: Concept, Human Factors, Manufacturing, Color Quality Control, and Point of Sale.

Figure 1 illustrates the fishbone developed for the causes of variability in the control of color within the
supply chain. The diagram contains some of the most important factors and procedures that may lead to
variability in the communication of color. It is important to bear in mind that a complete fishbone diagram
can be developed for each sub-factor. Detailed discussion of each of the main factors relating to Figure 1
above was reported recently [9].


National Textile Center Annual Report: November 2007
NTC Project: C04 NS11
4
Manufacturing Color Quality Control
Type of
instrument
Instrumental
Assessment
Cal ibration &
specificati on
Geometry
Sampl e preparation
Software &
equations
Accuracy &
repeatability
Key
variables
in the control of
color in the
supply chain
Lab trials
Machinery
Dyes and
pigments
Recipes
Control parameters
Substrate
Concept
Technical
Specifications
Cost Material s
Original Design
Hardware Software
Market place & trends
Human Factor
Trends
Market
Color Management
System
Equipment
Color production system CAD
Point of Sale
Store
Floor plan
Display &
surrounds
Handli ng &
storage
Rotation of
samples
Lighting
Non Store
Television
Internet
Catalog
Cal ibration &
settings
Resolution
Color
gamut
Color production
system
Design
Monitor
Browser
Paper
Vi ewing
conditions
Print
quality
Design
Human Factor
Viewing Conditions
Adaptation mechanisms Surround
Lighting
Appearance
phenomena
Psychological
factors
Culture
Past Experience
Emotions Training
Physiological
Acquired col or
defici ency Fatigue
Gender
Inherited color
defici ency
Age
Standards
Digital
data
Atlases
Substrate
Specifications
Bulk Production
(batch & continuous)
Dyei ng
Finishing
Preparation
Printing
See l ab
trials
See l ab
trials
See lab
tri al s
Color definition
& color difference
Tol erances
Visual
Assessment
Accuracy &
repeatability
Geometry
Booth &
specifi cations
Human
factor
Material type
Environmental
conditions
Tolerances
Observers
Q.C Personnel
Operator & Manufacturer
Customer
Buyer
Designer
Manufacturing Color Quality Control
Type of
instrument
Instrumental
Assessment
Cal ibration &
specificati on
Geometry
Sampl e preparation
Software &
equations
Accuracy &
repeatability
Instrumental
Assessment
Cal ibration &
specificati on
Geometry
Sampl e preparation
Software &
equations
Accuracy &
repeatability
Key
variables
in the control of
color in the
supply chain
Lab trials
Machinery
Dyes and
pigments
Recipes
Control parameters
Substrate
Lab trials
Machinery
Dyes and
pigments
Recipes
Control parameters
Substrate
Concept
Technical
Specifications
Cost Material s
Technical
Specifications
Cost Material s
Original Design
Hardware Software
Market place & trends
Human Factor
Trends
Market
Color Management
System
Equipment
Color production system CAD
Original Design
Hardware Software
Market place & trends
Human Factor
Trends
Market
Color Management
System
Equipment
Color production system CAD
Point of Sale
Store
Floor plan
Display &
surrounds
Handli ng &
storage
Rotation of
samples
Lighting
Store
Floor plan
Display &
surrounds
Handli ng &
storage
Rotation of
samples
Lighting
Non Store
Television
Internet
Catalog
Cal ibration &
settings
Resolution
Color
gamut
Color production
system
Design
Monitor
Browser
Paper
Vi ewing
conditions
Print
quality
Design
Human Factor
Viewing Conditions
Adaptation mechanisms Surround
Lighting
Appearance
phenomena
Viewing Conditions
Adaptation mechanisms Surround
Lighting
Appearance
phenomena
Psychological
factors
Culture
Past Experience
Emotions Training
Psychological
factors
Culture
Past Experience
Emotions Training
Physiological
Acquired col or
defici ency Fatigue
Gender
Inherited color
defici ency
Age
Physiological
Acquired col or
defici ency Fatigue
Gender
Inherited color
defici ency
Age
Standards
Digital
data
Atlases
Substrate
Specifications
Standards
Digital
data
Atlases
Substrate
Specifications
Bulk Production
(batch & continuous)
Dyei ng
Finishing
Preparation
Printing
See l ab
trials
See l ab
trials
See lab
tri al s
Bulk Production
(batch & continuous)
Dyei ng
Finishing
Preparation
Printing
See l ab
trials
See l ab
trials
See lab
tri al s
Color definition
& color difference
Tol erances
Visual
Assessment
Accuracy &
repeatability
Geometry
Booth &
specifi cations
Human
factor
Material type
Environmental
conditions
Tolerances
Visual
Assessment
Accuracy &
repeatability
Geometry
Booth &
specifi cations
Human
factor
Material type
Environmental
conditions
Tolerances
Observers
Q.C Personnel
Operator & Manufacturer
Customer
Buyer
Designer
Observers
Q.C Personnel
Operator & Manufacturer
Customer
Buyer
Designer



Figure 1. Variables in the Control of Color in the Textile Supply Chain.


-50
-40
-30
-20
-10
0
10
20
30
40
50
-50 -40 -30 -20 -10 0 10 20 30 40 50
a
b
















Figure 2. The location of dyed color standards in the CIE a*b* plane.

Visual Assessment of Color Difference
Nave versus Expert Observers
For the purposes of this experiment nine sets each consisting of a standard with six samples, dyed with
disperse dyes on unbrightened plain weave spun polyester fabric were used. Figure 2 shows the location of
each sample in a CIE a* b* plane. Each sample was cut to precise 22 dimensions and mounted onto
custom manufactured plastic holders. The sample mountings used precision cut PVC as backing and all the
components to the surround were uniformly spray painted to a L* of 74 which is approximately equivalent
National Textile Center Annual Report: November 2007
NTC Project: C04 NS11
5
to Munsell N 7.25. Each sample mounting could slide in a bar on a custom designed display easel, as
shown in Figure 3. With this setup sharp dividing lines were produced with no shadows. Test samples
were measured spectrophotometrically 3 times using a Datacolor International SF600 spectrophotometer
with the following setup: specular included, UV included, illuminant D
65
and 10 degree standard observer.
Each measurement was based on an average of 4 readings. The average of 3 measurements was then taken.
The sample pairs had an average E CMC
(2:1)
of 1.74, with a color difference range of 0.48-4.52 which
varied in lightness, chroma and hue.

Sample viewing
The easel was viewed at a 45 angle and was located in a Macbeth Spectralite III standard lightbox,
illuminated with a filtered tungsten daylight simulating lamp with a correlated color temperature of
6500100K and constant illuminance of approximately 1400 lx in the middle of the display board. All
extraneous light was eliminated. The light source was carefully controlled during the experiment in order to
diminish variability at constant room temperature.













Figure 3. Custom set up for visual assessment of color difference using an AATCC gray scale.

The observers wore a mid-grey lab coat and a pair of mid-grey gloves. The samples were placed by the
experimenter who also wore a mid-grey laboratory coat. At the beginning of the experiment the observers
eyes were adapted for 3 minutes by looking at the empty viewing booth during which time the experiment
was explained to the observer.

Psychophysical method
In this experiment an AATCC Gray Scale for visual assessment of change of shade was used as a guide for
assessing the perceptual differences in color. For each sample pair the question asked was: Which grey
scale difference is in closest agreement with the difference between the displayed sample pair? The result
can be between two steps, such as 3-4. In the current experiment 50 observers participated, 25 nave
(mostly students of North Carolina State University, tested for normal color perception using the Neitz test,
of which 11 were females and 14 were males) and 25 expert observers (industrial-based colorists from the
U.S. textile industry, including 10 females and 15 males). Each observer sat in front of the box so that
he/she could move the reference grey scale freely. Each nave observer assessed the differences 3 times on
separate days. Each expert observer assessed the sample set once.

Results
Visual replication using standard AATCC Gray Scale
Following a literature review of visual experiments to assess color difference, the methodology for the
visual replication experiment has been finalized. The currently accepted-as-valid visual datasets that have
been used to produce color difference formula have used various:
substrates (e.g., wool fabric, sewing thread, and paint chips);
light sources (e.g., different daylight simulators);
numbers and type of observers: some naive (i.e., no previous experience), and some expert;
methods of judgment (e.g., geometric gray scale or gray scale anchor pair);
methods of data analysis.
National Textile Center Annual Report: November 2007
NTC Project: C04 NS11
6
All these factors, we believe, may lead to statistically significant variability in the dataset produced. This
kind of variability may limit the effectiveness of the color difference model upon which the visual data set
is based. Furthermore, none of the five most referenced visual color difference experiments were fully
replicated to show reliability of the data. For these reasons, we believe it is imperative to establish a highly
controlled replicated visual experiment, with replicated data obtained from three major regions of the world
(US, Europe and Asia).
A total of 3100 assessments were made using 31 sample pairs; this represents a subset of the actual samples
to be used in the international visual replication experiment. The AATCC gray scale for color change
consists of 9 steps of color difference determined by the CIE 1976 L*a*b* (CIELAB) formula, and the
values reported by observes were converted to visual difference, DV, via a 3
rd
degree polynomial curve
fitting. Figure 4 shows the average results in grey scale grade units for the three repetitions carried out by
the nave observers and the average grade for expert observers. The average visual difference for each pair
was compared for each trial as well as the experts ratings. A paired t-test, results of which are summarized
in Table 1, was used to evaluate any statistical difference between each trial among the nave observers.

The average perceived visual difference for the expert observers was about 9% higher than that for the
average of the three trials carried out by the nave observers. This suggests that expert observers on average
assess chromatic differences to be perceptually larger compared to nave observers. It is also indicates that
the long term experience of observers in making color difference judgments may play a significant role in
visual datasets of the kind employed for the development of color difference formulas. The data infer that
expert observers are more strict in their assessment of color differences.
.
0
2
4
6
8
10
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Pai r Number
D
V
Trial 1
Trial 2
Trial 3
Experts

Figure 4. DV for the visual assessments for nave (three trials) and expert observers.

Table 1. Summary statistics for assessments carried out by nave assessors.
Group t P Significance
Trial 1 vs. Trial 2 -1.69 0.1016 No significant difference at =0.05
Trial 2 vs. Trial 3 3.94 0.0004 Significant difference at =0.05
Trial 1 vs. Trial 3 0.62 0.5428 No significant difference at =0.05

Results of a paired t-test between the first and the third trial conducted by nave observers indicate that, for
the method employed, the repetition of the visual observations does not significantly affect the assessment
of small color differences. However, the method employed is likely to a play a dominant role in the
significance of training. For example, Mangine et al. found a strong training effect using the same dyed
sample as the present work, but a different experimental procedure and a different observer group. The
Mangine experiment used a paired comparison test in which training nave observers via repeat
assessments produced more consistent results [29]. Additionally, the results of Wilcoxon signed rank test
were in agreement with the results of the t-test shown in Table 1.

Due to geographic constraints and availability, the expert observers performed the experiment only once.
Table 2 shows the results of paired t-tests between average data for each of the three nave observer trials
and the average experts assessments. A significant difference at a 95% confidence interval exists between
the nave and expert observers.
National Textile Center Annual Report: November 2007
NTC Project: C04 NS11
7

Table 2. Summary statistics for the nave observer trials vs. expert observers
Group t P Significance
Trial 1 vs. Experts -9.286 <0.0001 Statistically significant difference at =0.05
Trial 2 vs. Experts -6.64 <0.0001 Statistically significant difference at =0.05
Trial 3 vs. Experts -7.67 <0.0001 Statistically significant difference at =0.05

Hence, these data infer a statistical likelihood that the general population will tend to be more lenient in
color difference judgments in consumer settings than experts who assist in the development and quality
control of commercial products. The next stage in the analysis of variability in nave and expert observers
is to determine the underlying systematic causes that lead to variability.

Intra-group variation
To determine the variability of individual observers in comparison to the group average in each trial, intra-
group variability was calculated. Assuming that the average perceived color difference for each pair in each
group represents the true value, the intra-group variability, or accuracy, is the degree to which each
observers assessments of each pair agrees with the true value of such pair in that trial. Table 3 summarizes
the mean standard deviation as an indicator of the intra-group variation among observers for each trial.

Table 3. Summary of intra-group standard deviation for nave and expert observers
Observer Trial 1 Trial 2 Trial 3 Experts
Mean 1.54 1.68 1.42 1.90
Minimum 0.76 0.61 0.67 0.83
Maximum 2.94 5.43 3.92 3.65
Range 2.18 4.82 3.25 2.82

It may be anticipated that experts experienced in color difference judgments would, as a group, exhibit less
intra-group variability than an identically-sized group of observers with no such experience. However, this
was not the case with the experiment employed in this study. The results indicate that the mean standard
deviation for nave observers is smaller compared to that for expert observers for all trials. Since experts
employed in this study represented different sectors of the color industry, the higher intra-group variation
for experts may be due to an inherent bias resulting from varying professional constraints and requirements
when judging color differences, previous knowledge regarding color discrimination, the period of
professional experience, and perhaps even psychological pressure to perform as a strict judge of color.

Variance Component Analysis for Naive Observers
Table 5 includes the variance components analysis for the nave observers. Each entry in the column is a
numeric measure of variability.
Table 5. Variance components estimates for nave observers
Source Estimated Value
Variance due to observers 0.76
Variance due to pairs 2.30
Variance due to interaction 0.61
Variance due to method 1.72

The highest variance shown in Table 5 is due to pairs. This is expected since the color differences of all the
sample pairs presented to the observers were different. While the variance due to observers is higher than
that for the effect of repeated assessments (shown as interaction in Table 5), it is smaller than the variance
due to the method, which essentially comprises all other variables in the system. Hence, the choice of
visual method (including display system, choice of sample pairs, and the number of observers, repeats and
samples) should be a primary concern when conducting color difference experiments, in addition to other
factors such as the choice of observers.

National Textile Center Annual Report: November 2007
NTC Project: C04 NS11
8
Performance of Color Difference Formulas Against Each Visual Dataset
To test the practical relevance of nave observer training, PF/3 values, a composite measure of statistical
performance, for three color difference formulas [23, 27, 28] were tested against the averaged data for each of the
nave observer trials and the expert observer visual data. Figure 5 shows the consistent reduction in PF/3 (which
corresponds to increasing correlation) with each additional trial for the nave observers. In the case of trial three, the
PF/3 data between nave observers and experts is similar. This demonstrates the importance of completing
replications in the case of nave observers (expert observers may also require multiple repetitions). These data will
be compared to other methods employed so far, including a gap between the reference and test sample pairs, and
employment of a perceptually linear (not geometric) gray scale.
0
10
20
30
40
50
60
nave t r ial 1 nave t r ial 2 nave t r ial 3 exper t s
P
F
/
3
DE
DECMC
DE00

Figure 5. Graph of PF/3 for CIELAB DE, DE CMC, and CIEDE2000 for the average
nave observers and expert observers.

Assessment of small color difference for one blue color center
The current ISO and AATCC color difference formula is known to perform relatively poorly in the blue (and other)
regions of color space. The new CIE formula, CIEDE2000, was designed to improve in this region. Hence, we
conducted a comprehensive study of visual color difference around one critical blue center, and compared statistical
performance against the major color difference formulas, using a custom built jumbo gray scale (Figure 6b).
Figure 6a shows the effect of varying the lightness weighting factor, K
L
(l in the case of DE
CMC
), on the statistical
performance of each formula. Interestingly all formulas optimize around K
L
or l around 1-1.3, and CIEDE2000 and
BFD formulas have the lowest PF/3 values which means the best performance for this data. Hence, we find support
for the improved performance of CIEDE2000 (and BFD) against the existing standard formula in this region of
space. However, much more comprehensive data is required in different regions of space before a conclusion that
CIEDE2000 offers a significant overall improvement in performance over DE
CMC
. Furthermore, the data show the
importance of optimizing the variable factors in the various formulas to the particular visual assessment data.
KL
0.0 0.5 1.0 1.5 2.0 2.5 3.0
P
F
/
3
0
20
40
60
80
100
CIE94
CMC
BFD
CIEDE2000

(a) (b)

Figure 6. (a) Graph of comparison of the mean visual color difference for each sample pair assessed using the
conventional AATCC Gray Scale (Pilots 1-3 and Experts), and the custom built Large Scale. (b) Custom set up
for visual assessment of color difference using a large-scale gray scale.
National Textile Center Annual Report: November 2007.
NTC Project: C04 NS11
9

Area Lighting Variability in Retail Stores
We have measured the lighting variability of more than 20 stores that sell the products of several leading US retail
ompanies [e.g., 30], using our standardized method of area lighting measurement [31]. Recent data collected
ntrolled within and between department stores to a significant degree.

isual observer data pertaining to color perception, especially the perception of the magnitude of color difference.
rence models are based on visual assessment, and hence systematic and random errors in the visual
Statistically valid observer sets, sample size, repetitions and all other variables must be optimized
titions and these observers may be used effectively in
ISO
s and females, and the experimental setup for females


We conc e
sample set, e al methodology, and conduct ANOVA on
petitions and test for any significant training. Many of the visual datasets upon which current formulas have been
es, M.S. Thesis, North Carolina State University,
.
Geol Lee, Renzo Shamey, David Hinks, Warren Jasper, Assessement of Color Difference Formulae
ina Crdenas, Rolf Kuehni, Renzo Shamey, Unique Hue Stimulus Selection Using Munsell
] sus Experts in
riables in the Control of Color from Concept
] Brunescence and the Perception of Color
trolling Color Difference and Color Constancy for Multiple
rk at ASTM and ISO, AATCC-ISCC Industrial Color Solutions
he
0] Seung Geol Lee, Rolf Kuehni, Warren Jasper,
ction Sides of
Color, Kansas City, p 35-37, April 29-May 1, 2007.
c
shows that it is possible for lighting to be co

Summary
Fundamentally, the performance of any integrated color control program through the textile supply chain is based on
v
All color diffe
data will lead to errors and sub-optimization in modeling. Poor color difference modeling will negatively impact
potentially fatally digital color communication systems within a supply chain. While experiments under this
project will continue into 2008, we have contributed to the basic and applied science behind visual assessment of
color difference, by identifying and minimizing controllable key variables, although far more work is still necessary.

Some of the highlights of our findings [1-18] to date include:
Choice of observers is important to experimental methodology used in color difference assessment
Nave observer performance improves with repe
the development of color difference models
CIEDE2000 and BFD color difference formulas performed significantly better than the current
and AATCC formula, DE
CMC(2:1)
, for a comprehensive set of 67 blue samples around one color center
Unique hue choices vary widely for male
influences variability significantly
Yellowing of the human lens with age is not a statistically significant factor in variability in the
assessment of small color differences of textile samples.
lud from this work that a good protocol for visual assessment is to use a significantly large observer and
mploy at least three repetitions, a highly controlled visu
re
optimized have not employed this level of training/repeatability.

References
[1] Seung Geol Lee, Assessment of Metrics in Color Spac
2007
[2] Seung
Around One Blue Color Center, Color Research and Application, to be submitted Oct. 2007.
[3] David Hinks, L
Color Chips, J. Opt. Soc. Am. A, 24(10) (2007) 3371-3378.
[4 Lina M Cardenas, Renzo Shamey, David Hinks, Roger Woodard, Comparison of Nave Ver
the Assessment of Color Difference, to be submitted, Oct. 2007.
[5] Lina Maria Crdenas, Renzo Shamey, David Hinks, Key Va
to Consumer, to be submitted Oct. 2007.
[6 Andrew Treece, Lina Crdenas, David Hinks, Renzo Shamey,
Difference, to be submitted Oct., 2007.
[7] David Hinks and Renzo Shamey, Con
Substrates, AATCC-ISCC Industrial Color Solutions Symposium, Charlotte, NC, February 21-22, 2007.
[8] David Hinks, Review of On-going Wo
Symposium, Charlotte, NC, February 21-22, 2007.
[9] Renzo Shamey, David Hinks, Lina Maria Crdenas, Key Variables in the Control of Color Throughout t
Textile Supply Chain, Autex Conference, North Carolina State University, June 11-14, 2006.
[1 David Hinks, Renzo Shamey, Lina Crdenas,
Variability in Visual Small Color Difference Assessment: What It Means for Color Difference Formula
Performance, Inter-Society Color Council Annual Meeting, Bridging the Creative and Produ
National Textile Center Annual Report: November 2007.
NTC Project: C04 NS11
10
[11] L. Crdenas, D. Hinks, R. Shamey, R. Kuehni, W. Jasper and M. Gunay, Comparison of Nave and Expert
Observers in the Assessment of Small Color Differences between Textile Samples, Imaging Science and
Technologys 3
rd
European Conference in Color in Graphics, Imaging and Vision, Book of Papers, Leeds
University, 2006.
[12] R. Shamey, D. Hinks, L.M. Crdenas, Key Variables in the Control of Color in the Supply Chain, Autex
2006 World Textile Conference, North Carolina State University, Raleigh, June, 2006.
D. Hinks, K. Noor, R. Shamey, L. Crdenas, W. Jasper, N. Cassill and J.R. Aspland, Towards Global Color
Control in the Tex

[13]
tile Supply Chain: The Role of Lighting in Color Perception, AATCC Conference and
[14] s in Color, Student
[15]
lina State University, Raleigh, March 6-7, 2006.
[18] le Supply Chain, Ekotex International Dyeing
[19]
[20] Calculation of small color differences for acceptability. AATCC
[21]
tion. CIE Publication No. 142-
[23] elopment of CIE 2000 Colour Difference Formula: CIEDE2000,
40-350.
e CIE, San Diego, June 25-July 2, 2003.
ternational Conference and Exhibition, Greenville,
[26] ence Formulae,
[27] oc
132.
cy, PhD thesis, Ohio State University, 2005.
Foundation Symposium, From Design to Retailer: Bridging
1] Radiometric Measurement of

ow
Graduate
Exhibition, October 25-27, 2005, Boston, USA.
L. Crdenas and K. Hyde, Relationship between Unique Hues and Small Difference
Paper Competition, AATCC Conference and Exhibition, October 25-27, 2005, Boston, USA.
L. Crdenas, R. Clonts, D. Hinks, R. Shamey, Optimizing Color
Control, Automotive Industry Workshop, North Caro
[16] R. Clonts, L. Crdenas, R. Shamey, D. Hinks, Color: Moving Forward,
Automotive Industry Workshop, Raleigh, March 6-7, 2006.
[17] L. Cardenas, D. Hinks, R. Shamey, W. Jasper, N. Cassill, R. Aspland, Optimizing Color Control
throughout the Supply Chain, NTC Annual Forum, Hilton Head, SC, March 2006, Winner of the best
poster award in the Chemistry Category.
D. Hinks, Optimizing Color Control Throughout the Texti
Symposium, Sept. 28, 2005, Istanbul, Turkey.
J.R. Aspland and P. Shanbhag, AATCC Review, Vol.4, 2004, pp.26-30.
AATCC Test Method 173-1998, CMC:
Technical Manual, pp. 311-315, 2005.
ISO International Standard 105-J03:1995.
[22] CIE Technical Report: Improvement to industrial colour-difference evalua
2001. Vienna: Central Bureau of the CIE, 2001.
M.R. Luo, Cui G, Rigg B., The Dev
Color Research and Application, 2001; 26:3
[24] J. Gay, and Hirschler R., Field Trials for CIEDE2000 Correlation of Visual and Instrumental
Pass/Fail Decisions in Industry, 25
th
Session of th
[25] K. Noor, Hinks D., Laidlaw A., Treadaway G., and Harold R., Comparison of the Performance of
CIEDE2000 and DE
CMC
, Book of Papers, AATCC In
SC, Sept. 10-12, 2003.
G.M. Gibert, Daga J.M., Gilabert E.J., Valldeperas J., Evaluation of Color Differ
Coloration Technol., 121 (3) (2005) 147-151.
M.R. Luo, and Rigg B. BFD (l:c) Colour Difference formula, Part II- Performance of the Formula, J S
Dyers Col, 1987;103, 126-
[28] S. Guan and Luo R. Investigation of Parametric Effects Using Small Colour Differences, Color
Research and Application, 1999; 24, 331-343.
[29] H. Mangine, Variability in Experimental Color Matching Conditions: Effects of Observers, Daylight
Simulators, and Color Inconstan
[30] D. Hinks S. Draper, Q. Che, M. Nakpathom, A. El-Shafei, and R. Connelly, Effect of Lighting Variability
on Color Difference, Book of Papers, AATCC
the Gap, June 20-23, 2000.
[3 D. Hinks, A. El-Shafei, S. Draper, Q. Che, M. Nakpathom, and R. Connelly,
Area Lighting Critical to Color Assessment in the Textile Industry, AATCC ICE, Winston Salem,
September 17-20, 2000.
Ackn ledgments:
students: Lina Cardenas, Fiber and Poly. Science Program, NCSU
Seung Geol Lee, MS Textile Chemistry, NCSU
ndergraduate student: U Andrew Treece
Associate: Postdoctoral Research Melih Gunay, NCSU.
Companies: D Datacolor International, Gap, Inc., Color Manager, Nike, In yStar, c., X-Rite.
Universities: r. Xin, niversity, UK. D Hong Kong Polytechnic, HK; Prof. Luo, Leeds U
National Textile Center Annual Report: November 2007.

You might also like