Estimating Spectral Reflectance from
Camera Responses Based on CIE XYZ
Tristimulus Values Under Multi-Illuminants
Xiandou Zhang,
1
* Qiang Wang,
1
Jincheng Li,
1
Xiaohui Zhou,
1
Yuechuan Yang,
1
Haisong Xu
2
1
School of Digital Media and Art Design, Hangzhou Dianzi University, Hangzhou 310018, China
2
State Key Laboratory of Modern Optical Instrumentation, College of Optical Science and Engineering, Zhejiang University, Hangzhou 310027, China
Received 22 July 2014; revised 13 March 2015; accepted 31 December 2015
Abstract: A new spectral reflectance estimation method
based on CIE XYZ values under multi-illuminants was
proposed to obtain multi-spectral images accurately by
using digital still cameras. CIE XYZ values under multi-
illuminants were initially predicted from raw RGB
responses by using a polynomial model with local training
samples. Then, spectral reflectance was constructed from
the predicted CIE XYZ values via the pseudo-inverse
method. Experimental results indicated that the new spec-
tral reflectance estimation method significantly outper-
formed the traditional colorimetric characterization
method without requiring extra training samples or
greatly increasing computational complexities.
V
C
2016 Wiley
Periodicals, Inc. Col Res Appl, 00, 000–000, 2016; Published Online 00
Month 2016 in Wiley Online Library (wileyonlinelibrary.com). DOI
10.1002/col.22037
Key words: reflectance; spectrum reconstruction; camera;
color reproduction
INTRODUCTION
Digital still cameras have been widely used to acquire
digital images in daily life because they are portable,
easy-to-use, and affordable. These cameras also have
potential for vast applications in color measurement as
well as in computer vision with state-of-the-art hardware
improvement and built-in image-enhancement software.
1
At present, colorimeters and spectrophotometers are
mainly used to measure color information of objects. The
objects to be measured should be solid, flat, and with a
certain size; contact is also required during the process.
The RGB responses of a camera can also record color
information without any limitation to the shape, form,
and size of an object. In addition, a camera does not
require contact with the object and can obtain the com-
plete color appearance information with pixel accuracy.
Digital cameras can also replace human eyes in computer
vision applications, such as object recognition in indus-
tries
2
and disease diagnosis
3,4
in medical treatments.
However, images captured by digital cameras depend on
the device and the light source. That is, digital cameras
acquire RGB responses that are specific to a device, and
the responses only describe the color information of the
object under a specific light source in its environment,
5
which restricts their applications in color measurement
and computer vision. Thus, if RGB responses are accu-
rately transformed into device-independent color values,
then a camera can be used to measure the color informa-
tion of an object. Extracting the light sources independent
color information from RGB responses is also helpful in
computer vision applications.
6
Colorimetric characterization models, such as the poly-
nomial model,
7,8
the back-propagation neural network,
9
and the look-up table (LUT),
10
have been proposed to
transform device-dependent RGB responses into device-
independent CIE XYZ values. However, CIE XYZ values
can only record color information under certain illumi-
nants, and they vary significantly with the spectral power
distribution of an illuminant. That is, CIE XYZ values are
*Correspondence to: Xiandou Zhang (e-mail: xiandouzhang@126.com)
Contract grant sponsor: National Natural Science Foundation of China; con-
tract grant number: 61205168; Contract grant sponsor: National Science and
Technology Support Program of China; contract grant number:
2012BAH91F03; Contract grant sponsor: Public Welfare Project of Zhe-
jiang Province; contract grant number: 2016C31G2040041.
V
C
2016 Wiley Periodicals, Inc.
Volume 00, Number 00, Month 2016 1