Approach of self-adaptive illumination

for optical imaging systems

Sheng Zhou,1,2 Jim Goodliffe,1 Hao Cai,2 Kui Zhou,2 and Xianglin Zhang2,*
1
School of Aeronautical and Automotive Engineering, Loughborough University, Leicestershire, LE11 3TU, UK
2
College of Materials Science and Engineering, Huazhong University of Science and Technology,
1037 Luoyu Road, Wuhan, China
*Corresponding author: hust_zxl@mail.hust.edu.cn

Received 14 May 2014; revised 28 June 2014; accepted 30 June 2014;

posted 3 July 2014 (Doc. ID 209767); published 7 August 2014

An optical imaging signal is vulnerable to undesired features such as ambient light illumination and
partial specular reflection from the target; the success of extracting target features from images depends
largely on appropriate design of illumination. This paper presents an approach for self-adaptive illumi-
nation for optical imaging systems. The proposed illumination system projects a reference image to a
target surface as an initial structured illumination, and then adjusts the projected image automatically
to compensate the negative influences of undesired features. After this self-adaptive control process,
undesired features would appear mostly invisible in the captured images. The signal-to-noise ratio would
be improved dramatically well before subsequent image processing. In the validation experiments,
several images with uniform brightness were offered as reference images; the captured images could
achieve high brightness uniformity, even when the target surface was uneven or was illuminated by
ambient light. In a further experiment of selective vessel illumination on a human palm, simulated vessel
regions were selectively illuminated. Undesired features, like palm prints, almost disappeared in the
images captured. © 2014 Optical Society of America
OCIS codes: (220.0220) Optical design and fabrication; (220.2945) Illumination design.
http://dx.doi.org/10.1364/AO.53.005205

1. Introduction undesired features cannot be neglected, that is,

With the development of optical imaging systems,
appropriate design of illumination has become more
and more important to keep pace with the highly
developed image acquisition techniques. In the first
place, uniformity is one of the most important factors
for illumination quality as it is essential to achieve a
good image signal. A lot of effort has been made to
achieve illumination uniformity, mostly by employ-
ing sophisticated optical components, such as lens
arrays [1], freeform lenses [2], and freeform reflectors
[3]. These components can help achieve uniform
illumination in particular situations, but not neces-
sarily in a practical optical imaging system where
1559-128X/14/235205-06$15.00/0
© 2014 Optical Society of America
ambient light, uneven target surfaces, inhomo-
geneous reflection–absorption properties of the tar-
get, and so on. For this reason, an illumination
system that suits the object to be illuminated is
desirable [4]. On the other hand, flexibility of illumi-
nation is required in areas like computer vision [4]
and image photoplethysmography (iPPG) [5], where
the illuminated objects are always complicated and
changeable. Several techniques have been developed
to realize flexible illumination. A great deal of work
has been done on applying structured illumination
to optical profilometers by LCD projectors [6–9].
A vision-measuring system equipped with self-
adapting illumination has been designed. With the
captured image of the illuminated target acting as
the feedback signal, the illumination intensity can
be adjusted automatically to improve the vision

10 August 2014 / Vol. 53, No. 23 / APPLIED OPTICS 5205

measurement accuracy [10]. Furthermore, in the
field of projection display systems, the concept of
camera-assisted projection has been proposed to dis-
play with real-time photometric adaptation to the
changing environment [11,12]. In other words, since
images are mainly the interaction result of illumina-
tion, the illuminated object, and the camera, full con-
trol of illumination would surely promote the ability
of extracting desired data from images.
In this paper, a self-adaptive illumination ap-
proach for optical imaging systems is proposed.
The proposed illumination system projects a refer-
ence image to a target surface as an initial structured
illumination, and then adjusts the projected image
automatically to compensate the negative influences
of undesired features. After this self-adaptive control
process, undesired features would appear mostly
invisible in the captured images. The signal-to-noise
ratio would be improved dramatically well before
subsequent image processing. A similar approach
has demonstrated considerable illumination distri-
bution control ability over the reflecting surface [13].
In the proposed illumination system, a shrunken im-
age of the reference image displayed on a LCD panel
is projected onto a target surface as an initial struc-
tured illumination. In the illumination adjustment
stage, an image of the illuminated target is captured Fig. 1. Optical layout: with the help of a Fresnel lens, patterned
light rays that transmit through the LCD panel will be focused
by a camera and acts as a feedback control signal.
onto the projection lens, reflected by the beam splitter, and illumi-
The illumination is then adjusted by refreshing nate the target plate. Images of the illuminated target are cap-
the image displayed on the LCD panel with a com- tured by a camera and sent to a computer as a feedback control
pensated image, which has been generated according signal in the illumination self-adaptation stage, or as an imaging
to the differences between the reference image and signal in the image data acquisition stage.
captured image. This self-adaptive process runs in
a loop until users are satisfied with the image ob-
tained. Until the self-adaptation process of the illu- algorithm, there is no longer a high illumination
mination system is accomplished, actual image data uniformity requirement for the light source.
acquisition for the optical imaging system could be
conducted. B. Control System
The program of the control system was developed in
Matlab. The control flow chart is shown in Fig. 2.
2. Method
The experimental setup of the proposed self-adaptive
illumination approach can be divided into two
sections, namely, an optical system and a control
system.
A. Optical System
The optical layout of the proposed illumination sys-
tem is shown in Fig. 1. Images on the LCD panel are
lit by the light source. In order to form bright illumi-
nation on the target surface, a Fresnel lens is used to
condense the light rays emitted from the LCD panel.
The LCD panel is kept beyond the double focus
length of the projection lens, so that a shrunken
image of the image displayed on the LCD panel is
projected on the target plate as the illumination.
In this way, rather high illumination intensity could
be reached despite the large light loss property of the
LCD panel. Additionally, the negative influence of
gaps between adjacent pixels on the illumination
could be minimized. Owing to the feedback control Fig. 2. Control flow chart.

5206 APPLIED OPTICS / Vol. 53, No. 23 / 10 August 2014

Users define a reference image (I 0 ) that they want to
capture from the target surface. The program sends
I 0 to the LCD panel as the first image displayed. The
first image is projected to the target surface and
captured by the camera. The camera then sends
the captured image (I 1 ) back to the computer as
the feedback control signal. The program registers
I 1 and forms a new registered image matrix (I 2 ),
which is in alignment with I 0 . In the ideal condition,
I 2
kI0 (k is a conversion ratio) because the average
brightness intensity of every corresponding area of
the image displayed on the LCD and captured by
the camera should be correlated linearly. However,
in this process, many factors will disrupt this linear
correlation, such as the influence of ambient light,
uneven illuminated surface, and inhomogeneous sur-
face reflection–absorption properties of the illumi-
nated target. To compensate for those disruptions,
the program compares I 0 and I 2 , determines the
differences, and sends a compensated image signal
(I 3 ) to the LCD panel, which adds compensation to
the former one displayed on the LCD panel. This
feedback process is repeated until the difference
between I 2 and I 0 is acceptable. The LCD panel will
keep displaying the last compensated image sent
from the computer as the final illumination pattern.
From then on, images captured by the camera would
be sent to the computer as imaging signals.
3. Experiment Conditions
In order to validate the feasibility of the proposed
self-adaptive method of illumination, a simple proto-
type of the proposed illumination system was built. A
3M 9080 overhead projector was refitted by replacing
the projection lens with a double convex lens with a
focal length of 100 mm and a diameter of 50 mm. A
LCD panel from a GNR TS700 monitor was disas-
sembled and placed on the stage surface of the pro-
jector. The distance between the LCD panel and the
projection lens was set at 400 mm so that a shrunken
image whose size is only one-third of the image
displayed on the LCD panel could be projected onto
the target surface. A Logitech low-cost webcam,
V-U0009, was used as the camera. For the control
system, a simple algorithm was tested. Display the
figure window of Matlab on the LCD panel, which
acts as an extended monitor of the computer; images
are shown in the figure window. With general
assembly accuracy, taking a photo of just the pro-
jected area on the target surface is unpractical; the
projected area in an image captured is more or less
a rotated or shifted version of the one on the target
surface, and so registrations of the images captured
are necessary. As long as the positions of the camera
and the target are settled, every captured image
shares the same transform matrix, so only one
manual registration would be needed for each sepa-
rate experiment. To get the transform matrix, a cal-
ibration process is conducted in the first place. An
image of a coordinate plane is displayed on the figure
window and is projected to the target surface at first.
The projected image is captured and sent to the com-
puter. The captured image is registered to match the
displayed image by manually matching the corre-
lated points in the two images; the transform matrix
of the registration is exported to the program. In the
feedback control process, I 0 is first shown on the fig-
ure window. The projected image is captured by the
camera and sent to the computer. The program regis-
ters the captured image with the transform matrix
obtained in the calibration process, compares I 2 with
I 0 , and replaces the image displayed on the LCD
panel by I 3 :


Io I2
I 3
I 0  t · meanI 0  · − ; (1)
meanI 0  meanI 2 
where t is the scale factor, which is set at 0.5 in all of
the following experiments.
The camera captures another image and the proc-
ess is repeated again. The image on the LCD panel
continues to be updated in this fashion until users
are satisfied with the quality of the image. After this
point, the last image displayed on the LCD panel is
used as the final illumination solution. It should be
noted that what was tested here is only a simple
algorithm to validate the principle of the proposed
self-adaptive illumination approach. In practice, ap-
plications of all other image-processing techniques
developed are encouraged to form better compen-
sated images to achieve higher self-adaptation ability.
4. Results and Discussion
A series of experiments was conducted to validate the
self-adaptation ability of the proposed illumination
approach.
In the first experiment, a flat, white piece of paper
was illuminated. The reference images were set as
I 0
K · ones480; 640; 3, which adapt to the resolu-
tion of the camera. K is the illumination intensity
factor. As shown in Fig. 3, images with high bright-
ness uniformity could be achieved with different val-
ues of k after 10 compensation cycles. The influences
of partial specular reflection from the illuminated
paper were compensated perfectly.
With conventional illumination systems, it is
almost impossible to capture an image with uniform
brightness when the target has a rather uneven sur-
face or is partly illuminated by external light. By self-
adaptive control, the proposed illumination system
makes it possible to some extent. In Fig. 4, the refer-
ence image is set as I 0
0.5 · ones480; 640; 3; it can
be seen that with compensations based on the differ-
ence between the captured image and the reference
image, the image displayed on the LCD panel will
refresh itself to meet certain requirements of the cap-
tured image in the proposed system. It can be seen
that images captured in different situations have
different brightness levels when using the same
reference image. This is because in order to reach
the best compensation ability, the control algorithm
used focused only on matching the brightness
10 August 2014 / Vol. 53, No. 23 / APPLIED OPTICS 5207

Fig. 3. Self-adaptation performance of illumination on a flat
white paper. (a) Reference image with k
0.5. (b)–(f) Images cap-
tured before the 1st, 2nd, 3rd, 4th, and 10th compensation cycle,
respectively, which together show how brightness uniformity im-
proved as the self-adaptation process progressed. (g) RGB bright-
ness distribution along the middle horizontal line of (b) and (f),
which shows that a high brightness uniformity has been reached
after illumination compensation. The brightness vibrations at the
extreme left and right positions are because these positions are
outside of the figure area displayed in the Matlab figure window.
(h) Captured image before the tenth compensation cycle with
k
0.3. (i) Captured image before the tenth compensation cycle
with k
0.8.
distribution of the captured images to that of the
reference image, not the mean brightness.
In addition, variances of the registered images of
those captured before and after self-adaptation of the
illumination under different conditions have been
compared and the results are shown in Table 1. The
variance value of an image represents its brightness
uniformity level. The variance value of the reference
image is 0. It can be seen that the variances of the
images captured on the same target reduced rapidly
after the self-adaptation process of the illumination
system. The final captured image shows high
brightness uniformity under all conditions tested.
In other words, the captured images match the
reference image well in brightness distribution.
5208 APPLIED OPTICS / Vol. 53, No. 23 / 10 August 2014
Fig. 4. Self-adaptation performance on (a) folded paper and
(b) with ambient light: Reference images were set as I0
0.5·
ones640; 480; 3. (a) Without any control, the captured image of
folded paper has many shadowy and bright areas. After 10 cycles
of self-adaptive compensation, uneven surface features have
almost disappeared in the captured image. From the tenth com-
pensated image, we can easily tell how the system adjusted the
pattern displayed on the LCD panel to achieve this. (b) Without
shade curtains, sunlight will make the captured images look partly
blue. After 10 cycles of self-adaptive compensation, the brightness
uniformity of the captured image improved a lot, even though a
small blue area still exists. From the compensated image, we
can see the system’s effort to compensate the blue area with more
yellow light input.
Traditional optical systems modify light patterns
through use of lenses, diaphragms, curved mirrors,
gratings, or optical fibers. They are typically limited
to simple patterns [14]. However, in biomedical imag-
ing, the targets to be illuminated are usually certain
tissues, like vessels or nerves. Illumination of unnec-
essary areas is not only useless, but also acts as noise
in the subsequent image processing stage. In an
attempt to get rid of this problem, several further ex-
periments have explored the potential of applying
the proposed illumination approach to realize selec-
tive illumination in biomedical imaging. A human
palm was illuminated by the proposed illumination
system. In the real situation of biomedical imaging,
our approach can only work on the assumption that
users know well the structure of the target and the
part that they want to selectively illuminated, so that
they can make a proper reference image that indi-
cates their preference of the image to be captured be-
forehand. Practically, editing the concerned features
on the image captured under ordinary illumination
might be a good way to make a reference image.
In this verification experiment, a sketch of the main
vessel net of the human palm acted as the reference
image. The quality of selective vessel illumination
looked good to the naked eye, but due to the poor
contrast ratio ability of the camera used, the cap-
tured images were of bad quality. Therefore, in the
 Table 1. Variance Change of Registered Images
k On Flat Paper
Condition 0.3 0.5 0.8
Before process 0.0121 0.0274 0.0168
After process 1.545E − 4 4.893E − 5 2.095E − 5
Fig. 5. Selective vessel illumination: a sketch that simulates the
real main vessel net on a human palm was applied as the reference
image and was projected to a human palm as the first captured
image shows. It can be seen in the tenth captured image that
after compensation, the simulated vessel region was clearly
illuminated. Undesired features, like palm print, became less
prominent.
improved experiment, an incarnadine background
was applied around the vessel region in the reference
image. The result shown in Fig. 5 demonstrates the
potential application of the proposed illumination
approach in the biomedical imaging field.
To realize self-adaptation performance, the pro-
posed approach applies image-processing techniques
to form a proper illumination pattern. Questions may
be asked about its advantage over direct image sig-
nal processing. The fact is that image-processing
techniques always get the demanded images at the
cost of a part of the raw image data. This drawback
of image-processing techniques may be acceptable
when the signal plays a major role in the data; but
it is not the same case when a weak signal is hidden
by a strong noise environment. In the proposed
approach, what is processed is not the image signal
itself, but the pattern of the illumination distribu-
tion. The illumination is applied based on the user’s
requirements and is adjusted to compensate for the
effects of undesired features on the target surface.
The final illumination is just like normal illumina-
tion, except that it is more personalized to the illumi-
nated target. High-quality image signals could be
captured directly by the camera without image
processing of the raw signal. For instance, in iPPG,
illumination could be applied to only the vessel
areas. In this way, the surrounding tissues would
not contribute to the image signal. At the same time,
undesired features, like palm prints, that contain
no useful signal could be weakened before image
acquisition.
5. Conclusion
In this paper, a self-adaptive illumination approach
for optical imaging systems has been proposed. Sev-
eral validation experiments have been conducted,
On Folded Paper (k
0.5) With Ambient Light (k
0.5)
0.0269 0.0193
1.147E − 4 1.421E − 4
and high performance of brightness distribution
controllability has been achieved on a flat paper
and a folded paper with and without ambient light.
Furthermore, vessel-selective illumination has been
achieved on a human palm, and primary undesired
feature compensation has been realized as well.
A lot of work can be done to apply the proposed
illumination approach to practical optical imaging
systems for better performance. Possible improve-
ments include better assembly accuracy, employing
high-quality industrial cameras, and applying better
image-processing techniques. It is also possible to
realize pixel-level illumination control through the
accurate calibration procedure proposed in [15]. It
is notable that the proposed illumination solution
also suffers from energy loss as other commercial
LCD products do; the brightness of the images cap-
tured was almost the same as that of the reference
images even though the light was condensed by
the Fresnel lens. However, it is believed that energy
loss would be largely reduced if a customized LCD
panel with no diffuser layer is used.
The authors would like to acknowledge the 7th
European Community Framework Programme for fi-
nancial support through a Marie Curie International
Research Staff Exchange Scheme (IRSES) Project
entitled “Micro-Multi-Material Manufacture to En-
able Multifunctional Miniaturised Devices (M6)”.
The authors would also like to thank Dr. Sijung
Hu for helpful discussions and support.
References
1. X. Deng, X. Liang, Z. Chen, W. Yu, and R. Ma, “Uniform illu-
mination of large targets using a lens array,” Appl. Opt. 25,
377–381 (1986).
2. Y. Ding, X. Liu, Z. R. Zheng, and P. F. Gu, “Freeform LED lens
for uniform illumination,” Opt. Express 16, 12958–12966
(2008).
3. Y. Ding, P. Gu, and Z. Zheng, “The freeform reflector for uni-
form rectangular illumination,” Jpn. J. Appl. Phys. 46, 7771
(2007).
4. Y. J. Roh, D. Lee, M. Kim, and H. Cho, “Visual inspection sys-
tem with flexible illumination and autofocusing,” Proc. SPIE
4902, 463–475 (2002).
5. Y. Sun, S. Hu, V. Azorin-Peris, R. Kalawsky, and S. Greenwald,
“Noncontact imaging photoplethysmography to effectively
access pulse rate variability,” J. Biomed. Opt. 18, 61205
(2013).
6. G. Sansoni, L. Biancardi, U. Minoni, and F. Docchio, “A novel,
adaptive system for 3-D optical profilometry using a liquid
crystal light projector,” IEEE Trans. Instrum. Meas. 43,
558–566 (1994).
7. G. Sansoni, F. Docchio, L. Biancardi, and U. Minoni, “Adaptive
3D optical profilometer using liquid crystal light projector,”
Proc. SPIE 2065, 230–236 (1994).
10 August 2014 / Vol. 53, No. 23 / APPLIED OPTICS 5209
 8. C. Zuo, Q. Chen, G. Gu, S. Feng, and F. Feng, “High-speed
three-dimensional profilometry for multiple objects with com-
plex shapes,” Opt. Express 20, 19493–19510 (2012).
9. S. S. Gorthi and P. Rastogi, “Fringe projection techniques:
whither we are?” Opt. Laser Eng. 48, 133–140 (2010).
10. Y. Qu, Z. Pu, Y. Wang, and G. Liu, “Design of self-adapting
illumination in the vision measuring system,” in 2003
International Conference on Machine Learning and Cyber-
netics (IEEE, 2003), pp. 2965–2969.
11. K. Fujii, M. D. Grossberg, and S. K. Nayar, “A projector–
camera system with real-time photometric adaptation for dy-
namic environments,” in IEEE Computer Society Conference
on Computer Vision and Pattern Recognition, CVPR 2005
(IEEE, 2005), pp. 814–821.
5210 APPLIED OPTICS / Vol. 53, No. 23 / 10 August 2014
12. S. K. Nayar, H. Peri, M. D. Grossberg, and P. N. Belhumeur,
“A projection system with radiometric compensation for
screen imperfections,” in ICCV Workshop on Projector–
Camera Systems (PROCAMS) (Citeseer, 2003).
13. P. Castellini, S. Cecchini, L. Stroppa, and N. Paone,
“Adaptive illumination through spatial modulation of light
intensity and image inversion,” Meas. Sci. Technol. 24,
055401 (2013).
14. C. Lutz, T. S. Otis, V. DeSars, S. Charpak, D. A. DiGregorio,
and V. Emiliani, “Holographic photolysis of caged neurotrans-
mitters,” Nat. Methods 5, 821–827 (2008).
15. P. Křížek, I. Raška, and G. M. Hagen, “Flexible structured
illumination microscope with a programmable illumination
array,” Opt. Express 20, 24585–24599 (2012).