COL 9(9), 093301(2011) CHINESE OPTICS LETTERS Septemb er 10, 2011
Camera-based model to predict the total difference between
effect coatings under directional illumination
Zhongning Huang (黄黄黄中中中宁宁宁)
1,2
, Haisong Xu (徐徐徐海海海松松松)
1∗
, and M. Ronnier Luo
2
1
State Key Laboratory of Modern Optical Instrumentation, Zhejiang University, Hangzhou 310027, China
2
Department of Color Science, University of Leeds, Leeds LS2 9JT, UK
∗
Corresp onding author: chsxu@zju.edu.cn
Received January 24, 2011; accepted March 31, 2011; posted online June 21, 2011
A camera-based model is established to predict the total difference for samples of metallic panels with
effect coatings under directional illumination, and the testing results indicate that the model can precisely
predict the total difference between samples with metallic coatings with satisfactory consistency to the
visual data. Due to the limited amount of testing samples, the model performance should be further
developed by increasing the training and testing samples.
OCIS codes: 330.1690, 330.1710, 330.1715, 330.1730.
doi: 10.3788/COL201109.093301.
Products with effect coatings have unique characteristics,
i.e., large change in appearance (color, glint, and coarse-
ness) under different viewing conditions. The traditional
methods of characterizing this kind of coatings can only
assess their color or the gloss level based on the standard
illumination and observation conditions recommended by
commission internationale de edairage (CIE) or interna-
tional standard organization (ISO) without judging the
glint or the coarseness effect of them, which is quite far
from the fact and cannot meet the demand of online
quality control in industries
[1,2]
. Hence, further studies
have become necessary and urgent due to the growing
number of products with effect coatings and the increas-
ing amount of metallic coatings produced by pigment
manufacturers
[3,4]
.
Metallic coatings are the most common and oldest
effect coatings used in modern industries
[3−6]
. In con-
trast to conventional solid coatings, the chromaticity and
lightness of metallic coatings strongly depend on illumi-
nation and viewing geometry, thus the metallic coatings
can accentuate the curved profile of an object, which
helps in attracting consumers via the amazing appear-
ance. Traditional instrumental methods for their charac-
terization always include multi-angle or multi-geometry
measurement and concentrate on the angle dependence
of the color. These methods, however, ignored texture
properties that might affect the appearance of effect
coatings heavily. For this reason, a camera-based model,
that could assess effective samples in terms of the total
difference, including color difference and glint difference,
was proposed in this study.
It has been found from the experience of industry appli-
cations that effective pigments affect not only the color
but also another aspect of coating appearance, namely
the visual texture
[5−7]
. Visual texture is the perceived
small-scale non-uniformity of the color when observed
within a distance of a meter or less. It is an important
property for its contribution to the overall appearance of
effect coatings. Under directional illumination, glint, be-
longing to the scop e of visual texture, is considered as the
most important attribute influencing the overall appear-
ance of effective pigments. It is defined as the tiny spot
that is strikingly brighter than its surrounding and only
visible under directional illumination conditions. The
glint effect would change when the observation geome-
try is altered, hence it is angle dependent. Glint value
is supposed to correlate with the local contrast between
the bright sparkle and its surrounding, and the amount of
sparkles as well. In this letter, a camera-based model to
predict the total difference for samples of metallic panels
with metallic coatings under directional illumination was
established.
The statistics of model performance was evaluated
in this study in terms of standardized residual sum of
squares (STRESS)
[8]
, which is calculated by
STRESS =
X
(∆E
i
− F ∆V
i
)
2
X
F
2
∆V
2
i
1/2
× 100, (1)
where i is the index, ∆E
i
and ∆V
i
are two groups of
data, and F always equals 1 since ∆E
i
and ∆V
i
have the
same unit in this study. Evidently, the STRESS value
will be 0 if the two groups of data are exactly the same.
The STRESS will increase with the rise in the difference
between the two groups.
The visual experiment was completed and illustrated
in the authors’ previous paper
[9]
, thus the focus in this
study is on the algorithm to predict the total difference
of sample pairs via images of physical samples based on a
digital camera. The framework of this model is given in
Fig. 1. The images of samples were taken by the digital
camera NIKON D80 in the same viewing geometry with
the visual experiment under directional illumination, as
shown in Fig. 2, to make sure that the samples could be
precisely represented in the captured images.
The camera was set with f -number (F ) of 6.3 and
exposure time of 1/40 s. The captured images were
3 872×2 592 pixels, with b oth the vertical and horizon-
tal resolution being 300 dpi. The central 1 000×1 000
pixels of the images, corresponding to the central area of
the samples, were chosen for further image analysis.
In a perfect imaging system, the camera response is
supposed to have a linear relationship with the incident
1671-7694/2011/093301(5) 093301-1
c
° 2011 Chinese Optics Letters
资源评论
