See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/285932328

Characterization of High-Fidelity color printing devices using Illuminant-

Independent approaches for color imaging application

Article · January 2008

CITATIONS READS

2 368

3 authors, including:

Mei-Chun Lo
Shih Hsin University
22 PUBLICATIONS   327 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Ministry of Science and Technology View project

All content following this page was uploaded by Mei-Chun Lo on 31 May 2016.

The user has requested enhancement of the downloaded file.

 Characterization of High-Fidelity Color Printing Devices Using
Illuminant-Independent Approaches for Color Imaging
Application
Mei-Chun Lo+, Chang-Lang Chen♥, and Tsung-Hsien Hsieh♥
+
Department of Information Management, Shih Hsin University, Taipei, Taiwan
♥
Department of Graphic Communication Arts, National Taiwan University of Arts, Taipei, Taiwan

of both the mean CIEDE2000 (i.e. ΔE*00) and the mean RMSE
Abstract were used. The results showed that two types of algorithms, used
The aim of this research was, optimally to derive a multi- in the printing process of 7-ink CMYKRGB, were successfully
spectrally illuminant-independent type of High-Fidelity (Hi- proposed. The mean ΔE*00 values are 1.04 and 0.79 in the
Fi )multi-color printing device characterization model, which forward and the reverse processes respectively for the MS type.
could reconstruct spectra of every color concerned in high- Also 0.00098 of the RMSE value was obtained for the reverse
dynamic-range of original images for an accurate color -matching transform of the MS type. It implies the spectral reflectance of
or -reproduction application. In this study, the Adobe RGB color every color in datasets was satisfactorily reconstructed.
space under the D50 illuminant condition was used as the reference Subsequently, following the preliminary study, by taking two
working space. To have the ability to adapt the change of light matters of “metamerism” and “wide-gamuts” in mind, feasible
sources, two works were preliminarily carried out. Firstly, an application modules of Hi-Fi multi-color process toward the
ideal type of CIEXYZ camera characterization model was derived spectral/complex color images was derived and proposed. Then,
using a spectral reconstruct method, based on basic vector under Sets of image-processing algorithms were implemented in the
the D50 condition.. Via the estimation and the reconstruction of the process of scanner-to-monitor, and then monitor-to-printer color
spectrum of every pixel in question, the XYZ under any light transform in the study. Eventually, to cross verify performances of
source considered could be obtained accordingly. Therefore, any models derived, a set of forced-choice paired comparison
Adobe RGB format of Hi-Fi complex color images could be psychophysical experiments carried out under a viewing phase of
transformed into the multispectral type of images via the CIEXYZ CIE D50 illuminant.
camera model derived. Secondarily, the characterization of 7-ink
CMYKRGB printing process, using FM screening technique, was Introduction
implemented and tested in terms of single stimulus. The superset of As well-known, it is not necessarily acceptable that a color
CMYKRGB i.e. Hi-fi color set) was divided into seven 4-ink reproduction has colorimetrically correct results, as metamerism
groupings, including KGCB, KBMR, KRYG, RKYM, GKYC, issue is concerned under all possible illumination conditions
BKMC, and also CMYK. Each subgroup contains three chromatic happened in real world. Therefore, it is the only actual solution to
inks and one black ink. Both the multispectral (MS, i.e. provide a detailed description of color properties of the color-
narrowband) and the broadband (BB) types of characterization sample surface considered in terms of its spectra, across color-
models, using 3rd-order and 2nd-order polynomial regression imaging media. That is to carry out spectral color reproduction and
equations respectively, were developed in this stage. They all spectrally define each color in terms of wavelength of its light.
applied singular value decomposition method (SVD) and Then, it can produce the physically identical effect for both an
constructed including both transforms of forward (i.e. fractional original and its reproduced image under every identical
dot area (FDA) to XYZ/L*a*b*) and reverse (i.e. XYZ/L*a*b* to circumstance (e.g. viewed in the same surroundings) if every pixel,
FDA) processes for each subgroup. The transform between device- with the same spatial coordinate in both original and reproduced
independent data and device-independent data was derived and images, has the same spectra.
defined via measuring a number of colors in the IT8.7/4 However, in addition to every-color spectra of images as
training/test target produced using each 4-ink grouping subset. An mentioned above, there is another important issue which needs to
approach of key component replacement (KCR), which is similar be addressed to achieve the objective of accurate color
to GCR (gray component replacement) applied in traditional reproduction. That is a mismatch problem that different imaging
CMYK printing process, was integrated to implement the BB type devices have variant color-gamuts. To cope with the tremendous
of a multi-ink color separation algorithm. Additionally, in order to development of optoelectronic age, the restriction of chromaticity
reconstruct the spectra and adaptively map to the corresponding gamut in the conventional 4-color CMYK printing process in
tone values (fractional dot area) for every color in test/training Graphic Arts can be practicably extended by the use of extra
datasets, the MS approach in the reverse process was optimized by colorants with the conventional CMYK primaries. Thus, via the
iterating the KCR components of BB algorithm in terms of the use of both the spectral approach and extra colorants, a High
measure of RMSE (root of mean square error). The predictive Fidelity of multi-color printing process with illuminant-
performances of 2 models derived were tested for each of independent characteristics can be optimally achieved.
subgroup in the IT8.7/4 datasets mentioned above. Two measures
 A schematic of Hi-Fi color gamut of CMYKRGB was studied condition (e.g. the D50 illuminat), could be rebuilt by using the
here, as shown in Figure 1. With different hues from standard inks Wiener method. The spectral reflectance of every pixel considered
of CMYK, extra colorants of RGB were used to offer the on images could be also accordingly computed.
opportunity to extend the printable gamut of standard process. This
idea [1] in Graphic Arts was considered, not only to advance the Printing Device Characterization
density range, the resolution and the rending of details; but also to
profit the representable gamut. Also the lighter or better hues of Modeling Color Behaviour and Subdivision
some saturated colors, as obtained in the conventional CMYK Approach for 7-ink CMYKRGB Printing Process
printing process, can be attained. Those colors are such as orange, Following the work derived previously [3,4], A heptatone (7-
violet, certain reds, blue, purple and certain greens, which cannot color) CMYKRGB process was proposed in this study, as shown
be reproduced by conventional CMYK printing. Indirectly, one in Figure 1, to extend the color gamut beyond what can be
can well imagine, then, it also augments the stability of tertiary achieved in the conventional CMYK printing system. The
colors. approach was carried out originally based on the scheme suggested
b* by Boll [5]. The derivation of characterization models of 7-ink
printing process was carried out by measuring a number of colors
Y
R in the IT8.7/4 test target produced using each subset of seven 4-ink
G groupings. These seven 4-ink groupings included CMYK, KGCB,
Extended gamut
KBMR, KRYG, RKYM, GKYC, and BKMC. Each subset
MY contained 3 chromatic inks and black ink. Six groups of 4-color
CY
a* subsets, except for CMYK subset, represent six adjacent and
K
M
overlapping subgamuts in the supergamut of a 7-ink CMYKRGB
color space. Each 4-ink set was separately characterized, as strictly
as a conventional CMYK ink set [4], to define the transform
C
between color (i.e. device-independent data) and ink (i.e. device-
CM Standard processing gamut
independent data) in its corresponding subgamut.. This approach
B resulted in the production of an inktable wherein every color of
Figure 1. The use of Hi-Fi colorants, R, G,, R to extend the gamut achieved
images reproduced was inked with a maximum of 4-inks.
with standard CMYK printing process
Broadband and Narrowband approaches
As mentioned earlier, two issues of “metamerism” and
Therefore, by keeping two kernels of both “metamerism” and “gamuts” were considered in this research. Therefore, two
“gamuts” in mind, the focus of this study was to derive a approaches of broadband and multispectral narrowband were
universally well-performing Hi-Fi multi-color printing device explored in the characterization process of 7-colors of CMYKRGB
characterization model. A feasible technique of FM screening printing device. Both multispectral and broadband approaches of
separation was used to process Hi-Fi complex color images by models numerically applied both a 3rd-order and a 2nd-order with
exploring a multi-spectral approach which is independent of 3rd-order polynomial regression equations, respectively. Those
illuminants. The conversion between device-dependent data (for models were all incorporated with a singular-value decomposition
instance CMY or CMYK) and device-independent data (i.e. CIE (SVD) technique [6]; and each carried out both a forward and a
XYZ/L*a*b*) hence, could be determined via the printing device reverse transform processes. The forward process maps the device-
characterization model. Here, the characterization was carried out dependent data (i.e. FADs, Fractional Dot Areas of four primary
via the provision of printing training data set of IT8.7/4. Finally it inks for a color considered in each subset tested) to their device-
could be satisfactorily applied in the process of cross-media colour independent values (i.e. CIEXYZ, CIELAB, or CIELCH); while
transform of Hi-Fi complex color images. the reverse process transforms device-independent values (i.e.
However, there was a necessity to have multispectral types of CIEXYZ, CIELAB, or CIELCH) into device-dependent data i.e.
complex color images as original input images in the process of FDAs). The broadband type of 2nd-SVD model (referred as 2nd-
FM screening separation. Consequently, a simulated multi- SVD-BB later) was derived from previous work [4, 7-8]. Also, by
spectrally performing camera characterization model was also following a work derived previously, a KCR (Key Component
derived in this research. It is an ideal type of CIEXYZ camera with Replacement) algorithm which is similar to the GCR (Gray
three-sensors of the 1931 CIEXYZ color-matching functions. An Component Replacement) technique in theory was also
adaptive method of basic vector was firstly carried out, via an implemented in each of subgamuts in this work for the 3rd-SVD-
SVD (Singular Value Decomposition) method. It was used to MS model. The former 4 of seven 4-ink subsets mentioned above
estimate the spectral radiances of tested color objects. Then, a set all had the key component of black; whereas the latter 3 of 7
of coefficients was approximated by using the Winner approach [2] subsets (i.e. RKYM and GKYC, and BKMC) had different key
via the simulated CIEXYZ camera of CIEXYZ-type spectral components of red, green, and blue respectively. Therefore, The
sensitivity functions of sensors. Therefore, the spectral energy KCR refers, e.g. in RKYM and KRYG subgamuts, to reduce key
(radiance) of every pixel in question (on images with known components R and K respectively (which are produced using
profile of XYZ values; here used Adobe type of images) which the paired inks of M & Y, and R & G respectively), and substitute
camera sensor picks up with basis vectors under a known light them with colorimetrically corresponding equivalent amounts of
red (R) and black (K) inks respectively.
 As for the model developed based on the multispectral device separately. Therefore four characterization models were
approach, it was implemented using the 3rd-order polynomial implemented and tested here. Those will be recognized as 3rd-
regression equations (recognized as 3rd-SVD-MS later). Table 1 SVD-MS-4c & 3rd-SVD-MS-7c, and 2nd-SVD-BB-4c & 2nd-SVD-
only demonstrates the deriving process of polynomial form for the BB-7c, respectively for multispectral narrowband and wideband
3rd-SVD-MS model. types later. As mentioned above both forward and reverse
transforms were implemented for each characterization model
rd
Table 1: The 3 -SVD Polynomial algorithm, used in the derived. Models’ performances were evaluated in terms of two
derivation of 7-ink printing device characterization model for the measures of Average (i.e. mean E00) and Max (i.e. maximum E00),
process of forward transform tested using their corresponding test targets (E00 is color difference
Orders Parameters Matrix Length of CIEDE2000). E00 was calculated between XYZ values of the
predicted color-patch and those of the original target color-patch in
3rd-SVD Rλ1 , Rλ 2 , 3
(∑Hi4 ) + 1 = 35 question. (The computational procedures used in both forward and
Equation Rλ 3 , Rλ 4 i =0 reverse transforms for 3rd-SVD-MS model, based on the
multispectral approach can be found in the previous paper [3]).
Polynomial Equations In the reverse transform of the spectrally-structured 3rd-SVD-
C 36 C 36 + C 25

∑a
j =1
j 3 rd ( R λ 1 , R λ 2 , R λ 3 , R λ 4 ) + ∑a j 2 nd ( R λ 1 , R λ 2 , R λ 3 , R λ 4 ) + MS model as shown in Figure 3, the same algorithm, used in the
reverse process of 2nd-SVD-BB based on the broadband approach
j = C 36 +1

C 36 + C 25 + C14 was integrated into the forward transform of 3rd-SVD-MS model,

∑a j 1 st ( R λ 1 , R λ 2 , R λ 3 , R λ 4 ) + 1 to iteratively look for the optimal solution of printing device-
j = C 36 + C 25 +1
dependent data (e.g. KRYG FDAs). In Figure 3, the Rλ of a target
Note: e.g. for the subset of KRGB, the Rλ1, Rλ2, Rλ3, and Rλ4 are the spectral color are first entered and then converted into XYZ values and CIE
reflectance for the corresponding FDAs of black, red, green, and blue inks 1976 a, b hue angle, hab under the illumination condition (e.g. D50
respectively. Then, the Rλ4c is the resulted spectral reflectance obtained from the here). The subset of 4-ink grouping, used to producing the target
overprint of KRGB inks in question. Here H denotes the rule of “Combination with color, is then determined via LUT of hue (through hab). The XYZ
Repetition”. values are also transformed into (Dr, Dg, Db)4c using the log density
functions given in equation (1). The next step is iteratively to
Experimental and Methodology predict the optimal amount of key component and then the FDA of
every ink used will be determined. In the optimization of iteration
process, every possible solution of considered subset’s (e.g.
KRYG) FDAs is input into the forward transform of 3rd-SVD-MS
model to reconstruct the spectral reflectance of target color of
interest in KRYG subgamut. Therefore, the best solution of
subset’s FDAs would be determined in terms of the E00 and also
the measure of RMSE (root of mean square error), The RMSE was
computed between the iteratively recovered (i.e. predicted) spectra
and original reference ones, of every color in the test target, in the
36-dimensional (i.e. 36 spectral wavelengths) space.
Xo Yo Zo
Dr = log , Dg = log , Db = log (1)
X Y Z

Figure 2. Produced IT8.7/4 CMYKRGB Test Target

Multispectral Characterization of Camera Model
A spectral reconstruct method, based on basis vectors, was
applied for the characterization of an ideally simulated CIEXYZ
type of camera in this study. The surface spectral radiance of an
A printing device selected was Heidelberg Speedmaster press
object under a specific illumination condition (e.g. D50 in this
using 7-color printing process. A printing characterization data set
study) measured by spectrophotometer ranged in 400-700nm at
of IT8.7/4 (with 1617 colors) was used in this study as the training
interval of 10nm, would produce 31 data-points. Therefore, for the
and test target, and generated for each of the sets of 4-ink grouping
GretagMacbeth ColorChecker SG (i.e. CCSG) with 140 color
in the 7-ink CMYKRGB printing process (K+E 918 Process Color
patches considered in this research, it would consist of a 31x140
plus PantoneRGB). A GretagMacbeth SpectraScan
matrix of color signal data. This color signal matrix can be
spectrophotometer was used in the measurement, performed at
composed of both basis matrix and coefficients matrix, shown as
45/0 geometry of illuminating and viewing using CIE 1931 20
equation (2).
observer. Samples were taken over the range across the visible
spectrum 380-730 nm with a 10 nm interval. There different ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
⎢31× 140 ⎥ = ⎢31× 3⎥ (2)
colorimetric data (i.e. device-dependent data of CIE XYZ, CIE
⎢ ⎥ ⎢ ⎥ • ⎢⎢3 ×140 ⎥⎥
⎣⎢ ⎦⎥ ⎣⎢ ⎦⎥ ⎣⎢ ⎦⎥
LAB and CIE LCH) were obtained against the CIE D50 illuminants.
Additionally, later on, both 4-color CMYK and 7-color ColorChecker Basis matrix; Coefficients matrix
Color Signal one column one basis
CMYKRGB models were all included in the characterization of
both of multispectral narrowband and wideband types of printing
 Three steps of procedure were carried out in the camera Wiener method. Therefore, the original spectral reflection (i.e.
characterization. The first, by applying SVD, all the spectral basis radiance) of objects considered could be rebuilt by multiplication
vectors, used to reconstruct spectral radiances of objects under the of both basis vectors and specific coefficients obtained. So, finally,
D50 illumination condition was found. Here, as mentioned above, the characterization models, derived via spectral reconstruction
a GretagMacbeth CCSG was used as a training test target. process using the optimized set/number of basis vectors, could be
Subsequently, the coefficients matrix was obtained using the applied under various illumination conditions.

Forward Model

Subsets

BroadBand Multispectral Band

CMYK

KCGB 3-ink Key

process Component Look-Up Tables

KRMB

Look-Up Tables Rλ1, Rλ2, Rλ3, Rλ4

KRGY

RMYK DbR, DbY, DrG rd

3 -SVD
Regression
CGYK rd DrK, DgK, DbK
3 -SVD
Regression
CMBK Minimizing
nd RMSE
2 -SVD
Dr3C, Dg3C, Db3C
Regression
false
true
Determine
Reverse
Subset KCR Iteration Break
Dr4C, Dg4C, Db4C
Key = 0~100

LCH
(for Hue) XYZ Reflectance

Reverse Model
Forward Model Reverse Model

Figure 3. The procedures of the forward and the reverse models used in both the multispectral and broadband types for 7-ink CMYKRGB printing process

Table 2: Prediction performances, in terms of mean ΔE00, of four derived Models for transform processes of both the forward
(denoted as F) and the reverse (denoted as R) under D50 condition.
CIE00 Colour Difference of Models
Multispectral Type Broadband Type
nd nd rd rd
2 -SVD-MS-4c 2 -SVD-MS-7c 3 -SVD-BB-4c 3 -SVD-BB-7c
Models
F R F R F R F R
Max 6.29 4.16 7.59 9.30 6.07 5.39 9.40 14.56
Average 1.28 0.70 1.04 0.79 1.38 0.73 1.05 0.73
ΔE00 >6 Count 1 0 4 7 1 0 4 23
RMSE (Mean) 0.00145 0.00072 0.00125 0.00098 - - - -

An extensive analysis of color spectra of the full set of SVD was used in this study. The purpose of this analysis was to
Munsell Book Glossy Database of 1600 color patches [9] based on statistically look for the most efficient basis number, for a given
color set by using basic functions. In the beginning, five basis images (displayed on the LCD) and the corresponding
vectors, considered sufficiently representing the spectral reproduction images (produced using DCMs derived).
accurately enough, were firstly selected for the analysis. However,
from the previous study [2,10], results all showed that three
vectors are adequately enough to achieve 93% of the variance of
spectra (i.e. spectral radiance).
Additionally, the illuminant D50, as considered practically
used in the Graphic Arts, was hypothetical here to provide an ideal (a)GATF (b)Sofa (c)Magic
modality to build a known database of spectral radiance by the
(d)Ski (e)Boats
multiplication of its SPD and the spectral reflectance of database
of interest (e.g. Munsell Book Glossy Database). Therefore, to
simulate the human eyes, a non-real camera using a triad of
sensors which had an ideal CIEXYZ type of spectral color-
matching responses/functions was chosen in this study. It
estimated original RGB (i.e. XYZ values for the ideal CIEXYZ Figure 4. Images used to in the process of monitor-to-printer colour
type of camera) under the D50 illuminant, and further multi- transformation
spectrally to derive a universal performing camera
characterization model. As a consequence, multispectral
characterization results could be effectively carry out under A panel of 10 observers viewed a paired of reproductions
various viewing light sources or illuminants with different color randomly presented, and judged which of the two gave a better
temperatures. Additionally, a well-characterised Adobe type of match (i.e. colour fidelity) to an original image in question. They
EIZO-CG221 LCD (with ΔE00 of 0.59) was also used here as the then rated the colour-fidelity quality of each hardcopy image,
source imaging media. Therefore, a set of Adobe RGB format of against its corresponding original softcopy, on a category scale
color complex-images were used in this study as original input with seven category scales of ordinal values from 1 (“ exact
images for the 7-ink Hi-Fi printing process. Every-pixel spectra match” ) through 4 (“acceptable match”) to 7 (“awful match”).
(radiance/reflectance) (on the Adobe RGB format of images with Both every original softcopy and its two corresponding
known profile of XYZ values), under a known D50 illumination reproduced hardcopies were displayed side-by-side on gray
condition, were firstly estimated using the Wiener method. Then background fields with approximate 50 of L*. A viewing cabinet
the predicted spectral data were, hence, further used in the color with a light source of D50 simulator was used to display two
separation of 7-ink printing process mentioned above. hardcopies. The colour temperature of the monitor was set to be
D50 condition.
Cross-media Color Transform
While both the forward – CMYK/CMYKRGB to Rλ to
XYZ–, and the reverse – XYZ to Rλ to CMYK/CMYKRGB –
AdobeRGB
transforms are of interest to characterize a printer or a press, the
inverse transform is of higher importance when it is strongly CMYK Coated
needed to accurately render Hi-Fi color complex-images on the
considered printer or press. Therefore, as mentioned, in addition to ISO Coated
both multispectral narrowband and wideband types of 7-color
CMYKRGB models, both multispectral narrowband and CMYKRGB Coated
wideband types of 4-color CMYK models were also considered
derived in this study. Also, to cross verify performances of four
printer DCMs, four set of corresponding image processing Figure 5. The color-gamut of 7-ink CMYKRGB (produced on coated paper)
algorithms were developed to implement a cross-media (from comparied to those of Adobe RGB, ISO coated, and CMYK coated.
camera to monitor, and then to printer) color transform using
complex image.
Totally, a set of 5 images, as shown in Fig. 4, were tested and
rendered using each of image processing algorithms implemented Table 3: The gamut volume of 7-ink CMYKRGB (produced on
as mentioned above. These 5 images included GATF, Sofa, Magic, coated paper) compared with those of Adobe RGB, ISO coated,
Ski and Boats. Both softcopy Images displayed on the Adobe type and CMYK coated.
of EIZO-CG221 LCD and respective hard-copy images (produced Gamut Names Gamut Volume
using the DCM methods) were representations of the originals and
the reproductions respectively. Adobe RGB 1299,180 Units

Psychological Experiment ISO Coated 398,483 Units

Finally, a set of forced-choice paired comparison CMYK Coated 420,310 Units
psychophysical experiments [11] was carried out, to make
comparisons of colour appearance matching between the original CMYKRGB Coated 569,229 Units
 Results and Discussions Acknowledgement
Table 2 lists prediction performances of each printer DCMs, This research has mainly supported by the National Science
obtained using the test data set of IT8.7/4. It seems that the 3rd- Council of Taiwan for financial assistance.
SVD-MS-4c and the 3rd-SVD-MS-7c performed very similar. Also,
the wideband type of two 2nd-SVD-BB models gave reasonable References
results. However, as mentioned earlier, it is needed to concern two [1] V. Ostromoukhov, Chromaticity gamut enhancement by heptatone
issues of “metamerism” and “color-gamut” in this study for Hi-Fi multi-color printing, IS&T/SPIE 1993 International Symposium on
accurate color reproduction. So, by using the multispectral Electronic Imaging: Science & Technology, Proceedings Conf.
approach, the metamerism had been optimally solved out. As for Device-Independent Color Imaging and Imaging Systems Integration,
SPIE, 1909, 139-151. (1993).
the color-gamut issue, the color-gamut produced on coated paper
[2] W. Ge, L. Chang-jun, Z.. Yun-long, and M. R.. Luo, “Improvement
using the 7-ink printing process was also investigated in this study, in the Estimation of Reflectance Functions Generated Using the
via the comparisons with those of Adobe RGB, ISO coated, and Basis Vectors”, Photographic Science and Photochemistry, 23(5),
CMYK coated media. Both Figure 5 and Table 3 illustrate the 340-350. (2005)
comparison results. They clearly show that, except Adobe RGB, [3] M. C. Lo, C. L. Chen and T. H. Hsieh, Characterization of High-
the 7-ink CMYKRGB color printing provided a satisfactorily wide fidelity Color Printing Devices Based on both Multispectral and
color-gamut, and had approximately 35.43% augment to the Broadband Approaches, MCS 2007 Conference: The 9th
performance of 4-ink printing color gamut. Summarily, the International Symposium on Multispectral Colour Science and
multispectral type of 2nd-SVD-MS-7c performed the best. Application, 36-44. (2007)
Additionally, the results evaluated from psychological [4] M. C. Lo, R. Chiang, The Characterization Models for Multi-colored
CMYKRGB, TAGA Proceeding, 242-254. (1998)
experiments also strongly verified that the 2nd-SVD-MS-7c
[5] H. Boll, A Color to Colorant Transformation for a Seven Ink Process.
outperformed the others, and gave satisfactorily Hi-Fi rendering
IS&T’s Third Technical Symposium on Prepress, Proofing, &
quality under all viewing conditions. Printing, 3, 31-36. (1993)
[6] W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery,
Conclusions Numerical Recipes in C-The Art of Scientific Computing, 2nd edition,
The research began with by taking two kernels of the matters (Cambridge University Press, 1992).
of both “metamerism” and “gamuts” in mind, and intended to [7] M. C. Lo, J. C.Chiang, and L. Shi, Color Separation for 7-ink
derive a universally well-performing Hi-Fi multi-color printing Printing Using FM Screening, TAGA proceeding, 716-735. (1997).
device characterization model which could be used in the cross- [8] M. C., Lo, C. L. Chen, R. K. Perng, and T. H. Hsieh, The
media color transform for complex images. Therefore, the model Characterisation of Colour Printing Devices via Physical, Numerical
derived needs to: 1) solve out metamerism problem to achieve the and LUT Models, CGIV 2006, IS&T’s Third European Conference
on Color in Graphics, Imaging, and Vision, University of Leeds, UK,
optimally requirement of illuminant-independent, and 2) produce
95-99. (2006).
a color-gamut that could be close to those of displays (used for [9] J. Orava, The Reflectance Spectra of 1600 Glossy Munsell Color
soft-proofing with high-rendition quality) or real dyes. Chips.
Consequently, two types of printing characterization (http://spectral.joensuu.fi/databases/download/munsell_spec_glossy_
algorithms, used in the printing process of 7-ink CMYKRGB all.htm)
using FM screening technique, were successfully proposed. Those [10] M. C. Lo, C. L, Chen, and X. Y. Shen, Characterization of Digital
were 2nd-SVD-BB and 3rd-SVD-MS models, applying the 2nd- Camera Based on Spectral Estimation and Reconstruction, MCS
order and the 3rd-order with 2nd-order polynomial regression 2007 Conference: The 9th International Symposium on Multispectral
equations respectively, via the broadband and the multispectral Colour Science and Applications, 167-173. (2007)
approaches respectively. Satisfactorily, the broadband type could [11] C. J. Bartleson and F, Grum, Visual Measurement, (Academic Press,
Inc, Optical Radiation Measurement , 1984), 5: 455-467.
be used to perform a colorimetrically correct color-matching of
color reproduction, but would only under a specified illuminant
considered. Since the 7-ink (heptatone) gamut approaches those of
Author Biography
Mei-Chun Lo holds a Ph.D. (1995) from the Department of
the film and high-quality display, the broadband type could be
Computer Studies and also a M.Sc. (2001) in Multimedia and Internet
also used to mitigate the problem of mapping unpredictable colors Computing (2001) from Loughborough University in England, and M.Sc.
onto printable colors (i.e. gamut mapping). As for the at the Institute of Graphic Arts Technology(1985) from Chinese Culture
multispectral one, it could be utilized both, not only to extend University, Taiwan. From 2005 to now, she is currently a full professor of
color gamut in the Hi-Fi multi-color printing system, but also to the Department of Information Management at Shih Hsin University. She
optimally reconstruct the spectral reflectance of colors in question has been researching and developing web-based cross-media color
(to release the problem of color-matching dependent on management system for digital archiving application and Graphic Arts.
illuminants). She is member of the Chinese Association of Graphic Science
&Technology, the Graphic Arts Technology Foundation (Now is
PIA/GATF), the Graphic Arts Technical Association (TAGA), the Color
Association of Taiwan.

View publication stats