Professional Documents
Culture Documents
2014MTA Aninvestigationofpixelresonancephenomenon
2014MTA Aninvestigationofpixelresonancephenomenon
2014MTA Aninvestigationofpixelresonancephenomenon
net/publication/271740279
CITATIONS READS
0 844
7 authors, including:
All content following this page was uploaded by Hsiao-Chuan Liu on 13 October 2016.
H.<C. Liu
Institute of Biomedical Engineering, National Taiwan University, No.1, Sec.1, Jen-Ai Road
Taipei 100, Taiwan
e-mail: tcliu0615@gmail.com
1 Introduction
An object of recognition, detection and classification can be realized by human vision using a
series of complex procedures, such as optical image formation, nerve interaction and brain
activity. However, these procedures can also be achieved by a computer vision system with
specific algorithms or construction techniques. Unique superiority is presented in the two
different vision systems. For example, human vision can extract more details from a stationary
scene. On the contrary, computer vision has the ability to capture various defects from a
sampled image, for example, in circuit board detection [9, 15].
Empirically, for human vision, dissimilar comprehension of a stationary scene could be
obtained from one person’s viewpoint of the scene at different points-in-time or from multiple
people’s standpoints at the same time, which is called multiple interpretations [6]. Multiple
interpretations result from human pattern recognition with special predictive mechanisms [3,
7]. For example, when people observe a pattern, it is always accompanied by two different
assumptions that are open to multiple interpretations. The first assumption is the classification
of image intensity (gray level). According to the predictive mechanism, people always focus
on the brighter area or on the more colorful area of a pattern before they start to realize the
pattern. The second assumption is a mechanism of visual attention [18, 19, 24]. People focus
C.<M. Tiu
Lotung Poh-Ai Hospital, No.83, Nanchang St., Luodong Township, Yilan County 265, Taiwan
C.<M. Tiu
Ching Chyuan Hospital, No.80, Sec. 4, Yatan Rd., Daya Dist., Taichung City 428, Taiwan
Y.<C. Hsu
Department of Electrical Engineering, National Tsing-Hua University, 101, Section 2, Kuang-Fu Road,
Hsinchu 30013, Taiwan
e-mail: ychsu@mx.nthu.edu.tw
Multimed Tools Appl
more of their attention on a particular location in an image (i.e., focusing on a pretty girl in an
image of a pretty girl in a general rustic landscape).
Unfortunately, some people do not have the ability to capture correct information or to
realize an accurate color from a picture. People with this symptom, which we called color
vision deficiency, are also called dichromats. Generally, color vision deficiency can be divided
into three different types: protanopia, deuteranopia and tritanopia. When the colors connected
by a single line cannot be distinguished by a dichromat, this line is called the pseudo-
isochromatic line (confusion line). Therefore, the types of color vision deficiency can be
decided by a confusion line of the International Commission on Illumination (CIE) 1931 color
space [14, 20, 26]. Additionally, the confusion line converges to a point, called the co-punctal
point, of which there are three, one for each type of color vision deficiency, which correspond
to different locations on the CIE 1931 color space [13]. Finally, the type of color vision
deficiency can be easily defined by confusion lines and the co-punctal point.
Color vision deficiency has been studied for many years [20]. Color vision testing is an
important procedure used to diagnose color vision deficiency in clinics. One popular method
used to test color vision is the Ishihara pseudoisochromatic test. The Ishihara
pseudoisochromatic plate, also called the color blindness test plate, is convenient and a
traditional scheme used to diagnose whether people suffer from a color vision deficiency
[2]. However, assisting people with a color vision deficiency in precisely distinguishing colors
is a very hard task.
According to previous studies [1, 16], the foreground and background colors of the Ishihara
pseudoisochromatic test plate can be decided by confusion lines. Therefore, in this study, the
confusion line was utilized to simulate the Ishihara pseudoisochromatic test plate and the
vision of dichromats. The reason for simulating the Ishihara test plate was because creating a
synthesized pattern similar to the Ishihara pseudoisochromatic test plate is a special technique
that can be applied everywhere in the future. Although related techniques of creating a
synthesized pattern are common, these techniques are always utilized to only make the
Ishihara pseudoisochromatic test plate. Our special technique can be used to camouflage
anything, such as an army pillbox wall, the surface of a tank or regimentals. However, this
camouflage technique is not the main point of this study, and we will develop this technique
with an omni-directional discussion in another paper. In this paper, we just used this camou-
flage technique to synthesize some patterns we wanted. In addition, we applied a creative
concept, pixel resonance (PR) based on the principle of photon resonance in quantum
mechanics, combined with the multiple interpretation theory, to aid people with color vision
deficiency in perceiving meaningful objects from a variegated background.
In this paper, we have applied an algorithm for PR research to dichromats. The proposed
concept of PR is quite novel. The concept of pixel resonance comes from the principle of
stochastic resonance (SR) proposed by Benzi et al. in 1981. In 2009, we proposed the concept
of PR for application to intensity images. In this study, we applied the concept of PR to color
images with a variegated background that have some information embedded, that is, the
Ishihara pseudoisochromatic test plate. We attempted to generate serial ways to help dichro-
mats to capture the correct information from a pattern with a variegated background. The
purpose of this study is to generate the PR phenomenon on the color Ishihara test plate. In the
following sections, the principle of PR in gray level imaging will be introduced, and Monte
Carlo random walks will be utilized to simulate the PR phenomenon. Moreover, the photon
casting concept with a Gaussian distribution will be applied and presented as the position of
people’s vision upon first glance. Furthermore, the three steps necessary to achieve the
simulation of color vision deficiency: the simulated Ishihara pseudoisochromatic test plate,
the simulated vision of dichromats, and the use of the PR concept to assist dichromats in
Multimed Tools Appl
distinguishing the correct number from an image with a variegated background, will be
described in section 3. The PR phenomenon for color images and limitations of the algorithm
will be presented and discussed in the results and discussion sections, respectively.
Far from being decreased, the performance of a system is actually increased while an
unpredictable stochastic oscillation exists in a nonlinear system, and this phenomenon is
presented as stochastic resonance (SR) [11, 12]. The phenomenon of SR can be widely
observed in various nonlinear systems. For example, according to the theory of quantum
mechanics, SR can be generated when photons are adulterated with some noise, such as heat or
electricity, such that these photons can jump over the threshold. Moreover, the quantum barrier
can be easily opened under a certain probability. Therefore, the aspect mentioned above will be
utilized in the image processing field. Adding a particular noise to some images can improve a
person’s recognition of these images. In this study, multiple interpretations were proposed to
aid people with a color vision deficiency in recognizing some patterns in images with a
variegated background.
2.2 The elucidation of stochastic resonance from photons and energy to intensity imaging
Visual perception is an important procedure in the human body for collecting information from
a scene or an image by a random movement of the eyeballs. The collected information could
include the intensity, the contrast, a special object and so on. These elements are analogous to
the behavior of a photon according to the theory of quantum mechanics. For example, a pixel
and the intensity of the image are analogous to a photon and the energy gap, respectively.
Random movement of the eyeballs, by which people watch a scene and transmit the received
random signals from the eyeballs to the brain, are analogous to random casting of the incident
photon clustering with a certain energy distribution against the target. Therefore, the PR
concept can be explained using the photon-electron interaction in the principle of quantum
mechanics, and the definition of PR Rn(x,y) can be presented as follows:
0
E n ; when I x;y −E n < T
Rn ðx; yÞ ¼ ; n ¼ 1; 2; 3 ð1Þ
0; otherwise
where E′n denotes a random variable with a Gaussian distribution of gray levels, a mean En and
a variance σ2. Ix,y defines the energy gap of a bonded electron at the position (x, y) in an image.
T is the threshold for triggering a single pixel to resonance. n is the position of a pixel in an
image. According to the principle of quantum mechanics, if the energy difference between En
and Ix,y is close, the resonance probability is greater. The concept of PR is illustrated in Fig. 1.
Therefore, various responsive PR patterns can be obtained from an image such that the concept
of multiple interpretations can be explained.
2.3 The simulation of the pixel resonance concept by the Monte Carlo method
Human vision is usually affected by some particular positions or postures from a scene or an
image that attract people’s attention, and this physiological behavior is analogous to a spotlight
Multimed Tools Appl
flashing on the ground or a sort of zoom lens focusing on an object [18, 24]. As section 2.2
mentioned, the PR behavior can be explained by the SR phenomenon of photons in quantum
mechanics. To simulate this phenomenon, a Monte Carlo simulation of random walks was
utilized. The Monte Carlo method can simulate the light distribution interacting with the
boundaries [21, 22]. However, the Monte Carlo method is also a useful tool in the field of
computer graphics. In this study, the Monte Carlo simulation of random walks was exploited to
simulate the photon casting process. This process is analogous to how people prioritize and
understand certain parts of a scene or an image. A photon casting on position (x, y) of a 2-
dimensional image presents as a vector:
x
x¼ ð2Þ
y
This simulation is easily achieved, and the pseudo code was presented in the Appendix
section.
3.1 The fundamentals of the CIE XYZ, CIE xyY and RGB space
Colorimetry is the science used to measure colors and to quantify the colors and
luminance that people perceive. We can extract the color features to achieve visual
recognition or multiple interpretations in the human vision system. Therefore, color
standardization is an important procedure. One of the color spaces is CIE XYZ 1931,
created by the International Commission on Illumination (CIE) in 1931. The CIE
XYZ 1931 space defines the tristimulus values of a visual system by three primary
parameters: X, Y and Z. By normalizing the X, Y and Z values, the CIE xyY chromaticity
diagram can be obtained to represent the colors in a 2-dimensional space. In the CIR xyY space,
Y is presented as luminance.
In contrast, according to the principle of colorimetry, the CIE XYZ space can be
transformed into the CIE RGB space by linear transformation. In the CIE RGB space,
Multimed Tools Appl
the R, G and B values are set in the range of 0 to 1 to easily make a space transformation. The
linear transformation equation from the CIE XYZ space to the CIE RGB space is presented as
follows [23]:
2 3 2 32 3
R 3:2406 −1:5372 −0:4986 X
4 G 5 ¼ 4 −0:9689 1:8758 0:0415 54 Y 5 ð3Þ
B 0:0557 −0:2040 1:0570 Z
where X, Y and Z are normalized such that Y is equal to 1 in reference to white and that the
values of R, G and B outside of the range are ignored. The chromaticity coordinate is based on
three points: R: x=0.64, y=0.33, z=0.03; G: x=0.30, y=0.60, z=0.10; and B: x=0.15, y=0.06,
z=0.79. The white point in this transformation is D65 (x=0.3127, y=0.3290, z=0.3583). An
approximate gamma correction value of 2.2 was also included in this transformation.
3.2 The extension theory of pixel resonance from intensity imaging to color imaging
b a
θðx; yÞ ¼ tan−1 ; ¼ Ι ðx; yÞRGB→xyY −Ο ð5Þ
a b
where En′ is a Gaussian random variable with mean En and variance σ2. I(x,y)RGB→xyY is a
color pixel positioned at (x, y) in the CIE xyY chromaticity diagram, and O is a reference
center point on the polar coordinate. Generally, the D65 point was employed as a reference
point. The color system with the polar coordinate is illustrated in Fig. 2a. Figure 2b demon-
strates the results of PR in color imaging according to Eqs. (4) and (5). The upper left corner of
this image is a color image with the number 37 and a green background. The upper right corner
of the image is the intensity image transformed by Eq. (5) with normalization of the angles to a
range between 0 and 255. The results of including PR in a color image were achieved and are
presented in Fig. 2b. The resulting PR effect is significant when En and σ were set at 90 and
10, respectively.
Fig. 2 The color system with polar coordinate and the processed results
simulated Ishihara pseudoisochromatic test plate was created in this study in three
steps. First, a circle pattern was constructed. Second, color sampling based on the rule
of confusion lines was obtained. Third, the parameter of luminance was placed on the
simulated color test plate.
Observing the true Ishihara pseudoisochromatic test plate, we can understand that entire circles
on the pattern are not in contact anymore. To generate a similar color blindness test plate, a
simulated Ishihara pseudoisochromatic test plate including non-overlapping random circles
with different sizes and positions was created using five steps. Step 1: a pixel at the corner of a
rectangular region with width W and length L is taken as a starting point (Fig. 3a) and called the
pixel square corner (i, j). Step 2: a value S from U(a, b), a uniform random variable in the
interval (a, b), is generated as the edge length to make a random square with (i, j) as a corner of
a square. Step 3: if the square drawn from (i, j) to (i+S, j+S) does not overlap with other
existing squares, this random square will be generated; otherwise, this random square will be
discarded (shown in Fig. 3c to e). Step 4: move to the next pixel, set it as a square corner (i, j)
and repeat steps 2 to 4. Step 5: an inscribed circle is drawn in each generated square, and the
square is then removed (Fig. 3f to h). In this study, the parameters W and L are both set at 256,
and S is a random value generated from a set [a b], where a and b were 5 and 15, respectively.
Multimed Tools Appl
Each circle on the simulated Ishihara pseudoisochromatic test plate has to fill up the specific
color to generate a well camouflaged background and an ambiguous number when the circle
pattern is constructed. According to the characteristics of confusion lines, the specific colors
inside the circle are chosen using the concept of random sampling. We assumed that there were
two one-dimensional Gaussian distributions defined by the mean and the variance and that
these two Gaussian distributions were applied to the Cartesian coordinates along one of the
confusion lines on the CIE xyY chromaticity diagram. The sampling colors can be expressed
as a color set including the locations and numbers of the CIE xyY diagram. The distribution of
the sampling color set is illustrated in Fig. 4a, and the equation of color sampling was defined
as follows:
C i ¼ cik cik ∈N μi ; σ2 ð6Þ
where N(μi,σ2) is the Gaussian distribution with mean value μi and variance σ2, and cik is a
color position vector on the CIE xyY chromaticity diagram. The values k and i denote the
number of colors in a set and the ith color set, respectively. Therefore, a simulated Ishihara
Multimed Tools Appl
Fig. 4 The distribution of the sampling color set and the simulated Ishihara pseudoisochromatic test plates
pseudoisochromatic test plate without the parameter of luminance can be obtained from the
two color sets shown in Fig. 4b.
Although the simulated Ishihara pseudoisochromatic test plate has been created, the
luminance of this plate is uniform. To imitate a more realistic simulation of the
original Ishihara pseudoisochromatic test plate, the parameter of luminance needs to
be determined. However, the CIE xyY chromaticity diagram does not include the
information of luminance. Thus, the transformation from the CIE xyY space to the
RGB space of the color blindness plate cannot be achieved. Although we can rescale
the terminal values of the coordinates in the RGB color space, this adjustment will
lead to the loss of the parameter of luminance. To diminish this problem, a method to
adjust the luminance was proposed. Because the transformation of XYZ and RGB are
linear, the different coordinates of the RGB space can map to the XYZ space. The uniform
luminance value can be calculated from the YIQ model, as shown in Eq. (7) [10], and the
simulated result is illustrated in Fig. 4c.
Y ¼ 0:299R þ 0:587G þ 0:114B ð7Þ
However, the simulated Ishihara pseudoisochromatic test plate with uniform luminance is
still recognized by dichromats because of the similar intensity. To enhance the difficulties in
recognition (such that people with a color vision deficiency have more difficulty than people
Multimed Tools Appl
without a color vision deficiency in recognizing this plate) and to obtain a reliable testing
result, the concept of random luminance was developed. The mathematical formula is pre-
sented as follows:
xyY to RGB i i
b i xyY to RGB
Ci → C i ¼ bck ck → bck ð8Þ
Y
cxi ¼ bcxi ð10Þ
Y
The simulated result is demonstrated in Fig. 4d. In addition, we adulterated the blue
component of the Ishihara pseudoisochromatic test plate such that the number became an
ambiguous number that looked like 3 or 8 and was hard to decipher. The distribution of the
sampling color set with the blue component and corresponding blindness test plate are
illustrated in Fig. 4e and f, respectively.
Approaches for simulating color vision deficiency have been widely proposed [1, 2, 11, 16, 17,
21, 22, 27], but the algorithm Brettel et al. proposed is the most popular [4, 5, 25]. This method
can convert colors from the red, green and blue (RGB) color space to the LMS color
space. A simulator has already been created (a website named Vischeck) that used
Brettel’s algorithm to simulate color vision deficiency [5, 8]. The LMS color space is
the best method to represent the response of the three types of cones of the human
eye, but this color space will reduce by one dimension because of the lack of one
cone cell in dichromats. However, the LMS color space with just two dimensions can
still be transformed to the RGB color space for normal people to observe. In this
study, we will observe how the simulated plate appears to people with the three
different types of color vision deficiencies: protanopia, deuteranopia and tritanopia, by
using different confusion lines and co-punctal points.
For trichromats, the parameter of hue is similar to an angle of the polar coordinate.
Therefore, to simulate the dichromatic appearance of the world, the Cartesian coordi-
nate for the chromaticity diagram has to be transformed into the polar coordinate
combined with the different co-punctal points that accompany the particular color
vision deficiency. Later, the angle of the polar coordinate will be mapped back to the
original image to generate the different vision seen by people with color blindness.
Figure 5 illustrates the results of the simulation of the world appearance of people with the three
different color vision deficiencies. Figure 5a and e demonstrate the vision of normal people.
Figure 5a and e, b and f, c and g, and d to h with the synthesis technique using random
luminance method demonstrate the vision of normal people, protanopes, deuteranopes and
tritanopes, respectively.
Multimed Tools Appl
Fig. 5 The results of the simulation of the world appearance of people with the three different color vision
deficiencies
4.1 The simulated Ishihara pseudoisochromatic test plate and the simulated vision of normal
people versus people with color blindness
In this study, the simulated plate was set as 256 by 256 pixels. The circle diameters were
uniform random variables in the interval [5, 16], and the luminance of the RGB colors were
rescaled as a constant or converted to a Gaussian random variable. The simulated Ishihara
pseudoisochromatic test plates with the number 3 and with a different luminance were
manufactured using a confusion line and are represented in the first three images in Fig. 6.
On the other hand, the simulated Ishihara pseudoisochromatic test plates with an ambiguous
number 3 or 8 and with a different luminance were created using three different color sets and
are represented in the last three images in Fig. 6. Obviously, the number 3 in the simulated
plate with the rescaled RGB color is clearer than the number 3 in the plates with luminance.
The multiple interpretations of people with color blindness, protanopia, deuteranopia, and
tritanopia, were also simulated by using a confusion line, and these results are illustrated in
three images of Fig. 6b, c and d, respectively. We clearly find that reorganization of the number
for protanopes and deuteranopes is difficult, and only an obscure number might be observed in
the simulated plate with the rescaled RGB color. For tritanopes, the number 3 is easily
perceived. In another more complex case with an ambiguous number 3 or 8, the number 3
in the simulated plate with the rescaled RGB color can be identified by all types of dichromats.
On the contrary, the observation of the simulated plate with luminance demonstrated that
protanopes and deuteranopes can perceive just part of the number. However, for tritanopes, the
number 3 can be easily perceived. The three pictures of Fig. 6b, c and d, demonstrate the
simulated vision of people with color blindness of three different color sets. However, we can
easily imagine that a protanopic or deuteranopic person who is observing this simulated
Ishihara pseudoisochromatic test plate with their particular perception ability should see
images similar to Fig. 6b and c.
Multimed Tools Appl
Fig. 6 The simulated Ishihara pseudoisochromatic test plates with the number 3 of different input luminance
To achieve the application of the PR concept to color imaging, the transformation from the
Cartesian coordinates to the polar coordinates with different center points was performed on
Fig. 4f to obtain an intensity image that should look the same to normal people and people with
color vision deficiency. The center point was assigned to the white point (D65) for normal
people and to three different co-punctal points for the three corresponding types of color
Fig. 7 The intensity images illustrated by the polar coordinate transformation with different co-punctal points
Multimed Tools Appl
blindness. The intensity images illustrated by the polar coordinate transformation with differ-
ent co-punctal points are shown in Fig. 7.
Moreover, the phenomenon of PR illustrated in Fig. 7 was used to aid people with color
vision deficiency to improve their perception. Seven sampling energies were included in this
study to observe the PR phenomenon, and the standard deviation (SD) of the sampling energy
and the threshold were set at 5 and 10, respectively. Figure 8 demonstrates the results of the PR
phenomenon with seven different energies. It is clear that the PR phenomenon is significant for
Fig. 8 Implementation of a simulated dichromat’s vision for the PR of both normal people ant tritanopes
Multimed Tools Appl
normal people and tritanopes, but protanopes and deuteranopes see nothing when the sampling
energy is set at 20. This result is because the gray level of the number 3 in Fig. 7a and d is
approximately 30, but the gray level in Fig. 7b and c is approximately 100, which is too far
from the sampling energy.
For protanopes and deuteranopes, unfortunately, the PR phenomenon does not appear with
any sampling energies. Although a small PR phenomenon appeared when the sampling energy
was set at 100 or 120, it did not add any meaning to this resonance pattern. In contrast, when
the sampling energy was set at 60, the normal people and tritanopes saw the number 8 and the
number 3 with a dark foreground combined with a bright background. This result is because
the PR phenomenon appeared in the background in Fig. 7a and d with a background gray level
of approximately 50. The gray level of the number 3 in Fig. 7a and d is close to 20; therefore,
the letter in the simulated color test plate becomes a dark color, which is quite reasonable.
However, the number 8 does not have to be resonated in the tritanopic cases at all sampling
energies because the gray level of the background and the left side of the number 3 are too
close, as shown in Fig. 7d. The PR phenomenon used to resonate the number 8 is not a hard
take by a fine design. In fact, the colors of the original Ishihara pseudoisochromatic test plate
are meticulously calculated and assigned. In this study, the concept of multiple interpretations
was elaborated upon by a simulated dichromat’s vision, and the PR phenomenon was achieved
in normal people and tritanopes, for which a significant resonance pattern is shown in the
upper right corner of Fig. 8. In addition, there is a key point we need to explain clearly. We
believe that the vision of people with color vision deficiency is similar to the results of our
simulations because all crucial parameters are defined according to the confusion lines and co-
punctal points of the CIE chromaticity diagram. When we or people with color vision
deficiency gradually adjust the threshold value, usable information embedded in the picture
will slowly appear. Moreover, the color Ishihara test plate will turn out to be an intensity image
by pixel resonance processing. Therefore, people with color vision deficiency can obtain the
information they wanted using the pixel resonance phenomenon.
5 Conclusions
This study proposed using the principle of PR to interpret the concept of multiple interpreta-
tions of color blindness test patterns. The principle of PR is mainly derived from the SR theory.
A Monte Carlo simulation of random walks is a common method used to achieve the SR
conception by simulating an experiment of the photon casting process. This process can be
analogous to how people prioritize and understand certain parts of a scene or an image. The
concept of PR applied to intensity imaging was introduced and presented in Section 2. Next,
the extension theory of PR conception to color imaging was proposed in this study. In addition,
we created a simulated Ishihara pseudoisochromatic test plate using three steps presented in
Section 3. The visual simulation of dichromats and normal people was achieved by the use of
confusion lines and co-punctal points. Moreover, the application of the PR phenomenon to
dichromats on the simulated Ishihara pseudoisochromatic test plates was accomplished. In this
study, we can aid a tritanopic person in obtaining the correct information. Unfortunately, the
application of the PR phenomenon to protanopes and deuteranopes does not present a mean-
ingful improvement in obtaining the correct information. In future studies, more detailed
research will be proposed to extend the application of the PR phenomenon to protanopes and
deuteranopes. Moreover, in the future, the concept of PR will be applied in clinics to assist people
with color vision deficiency in recognizing the correct objects from a variegated background.
Multimed Tools Appl
Appendix
The photon casting process by Monte Carlo simulation of random walk was achieved by a
pseudo code, and shown as follows.
Pseudo code:
for each random casting pixel I(x,y) of the image
for n = 1 to N
Rn(x,y) = En
else-if
Rn (x,y) = 0
end-if
end-for
end-for
References
1. Anishchenko VS, Astakhov V, Neiman A (2007) Nonlinear dynamics of chaotic and stochastic systems:
tutorial and modern developments. Springer Verlag
2. Attarchi MS, Labbafinejad Y, Mohammadi S (2004) Occupational exposure to different levels of mixed
organic solvents and colour vision impairment. Neurotoxicol Teratol 32:558–562
3. Biedeman I (1987) Recognition-by-components: a theory of human image understanding. Psychol Rev 94:5–147
4. Birch J (1997) Efficiency of the Ishihara test for identifying red-green colour deficiency. Ophthalmic Physiol
Opt 17:403–408
5. Brettel H, Viénot F, Mollon J (1997) Computerized simulation of color appearance for dichromats. J Opt Soc
Am A 14:2647–2655
6. Cheng CM, Wu SJ (1994) Orthographic satiation and disorganization in Chinese. Adv Stud Chin Lang Proc
20:183–190
7. Dai H, Micheyl C (2010) Psychophysical reverse correlation with multiple response alternatives. J Exp
Psychol Hum Percept Perform 36:976–993
8. Dougherty B, Wade A, Vischeck. Available: http://www.vischeck.com
9. Forsyth DA, Ponce J (2002) Computer vision: a modern approach. Prentice Hall
10. Gonzalez RC, Woods RE (2002) digital image processing. Addison-Wesley Pub (Sd)
11. Hammersley JM, Handscomb DC (1975) Monte carlo methods. Taylor & Francis
12. Hsieh CW, Hsu YC, Jong TL (2009) Image interpretations based on quantum resonance concept.
Multimedia and Ubiquitous Engineering, Qingdao, pp 59–63
13. Hunt RWG (1968) Colour science: concepts and methods, quantitative data and formulas. J Mod Opt 15:197
14. Judd DB, Wyszecki G (1975) Color in business, science, and industry. THIRD EDITION, Wiley-Interscience
15. Kay KN, Naselaris T, Prenger RJ, Gallant JL (2008) Identifying natural images from human brain activity.
Nature 452:352–355
16. Lee DY, Honson M (2003) Chromatic variation of Ishihara diagnostic plates. Color Res Appl 28:267–276
17. Logvinenko AD, Beattie LL (2011) Partial hue-matching. Vis Res 8:6
18. Noudoost B, Chang MH, Steinmetz NA, Moore T (2010) Top-down control of visual attention. Curr Opin
Neurobiol 9:353–364
Multimed Tools Appl
19. Peelen MV, Fei-Fei L, Kastner S (2009) Neural mechanisms of rapid natural scene categorization in human
visual cortex. Nature 460:94–97
20. Pitt FHG (1944) The nature of normal trichromatic and dichromatic vision. Proc Roy Soc Lond B Biol Sci
132:101–117
21. Rubenstein RY (1981) Simulation and the Monte Carlo method. Wiley, New York
22. Szirmay-Kalos L (2000) Monte-Carlo methods in global illumination
23. Truchetet F, Laligant O (2008) Review of industrial applications of wavelet and multiresolution-based signal
and image processing. J Electron Imaging 17:031102
24. Verghese P (2001) Visual search and attention: a signal detection theory approach. Neuron 31:523–535
25. Vienot F, Brettel H (2000) Color display for dichromats. Proceedings of SPIE, San Jose, pp 199–207
26. Wright WD (1952) The characteristics of tritanopia. J Opt Soc Am 42:509–517
27. Zjakić I, Parac-Osterman D, Bates I (2011) New approach to metamerism measurement on halftone color
images. Measurement 44:1441–1447
Chi-Wen Hsieh is an associate professor in the Department of Electrical Engineering at National Chiayi
University, Taiwan since 2009. He received Ph.D. from the Department of Electrical Engineering, National
Tsing Hua University, Taiwan in 2007. He also received his MS degrees in Department of Physics from National
Tsing Hua Universit, Taiwan in 1993. His research interests include biomedical information, multimedia video
with langue learning and green energy technology.
Hsiao-Chuan Liu is a Ph.D student in Department of Biomedical Engineering at University of Southern California.
He received the MS degree in Institute of Biomedical Engineering at National Yang Ming University, Taipei,
Taiwan, in 2005. He is also working as the position of the research assistant started from 2007 to present in the
Department of Radiology at Taipei Veterans General Hospital. His research interests include biomedical imaging
processing, computer-aided diagnosis system, bone age studies and histopathologic research for breast cancer.
Multimed Tools Appl
Chih-Yen Chen was born in Taoyuan, Taiwan, in 1977. He received the MS degree from the
Department of Physics, at National Taiwan Normal University and Ph.D. from the Institute of
Biomedical Engineering, National Yang-Ming University in 2003 and 2010, respectively. Now, he is
an assistant researcher at the Instrument Technology Research Center, National Applied Research Labora-
tories, from 2011 to present. His research interests include image processing, computer-aided diagnosis, and
human machine intelligence.
Yi-Hong Chou is chief of Ultrasound Section at Taipei Veterans General Hospital and Professor of
Radiology at National Yang Ming University School of Medicine, Taipei, Taiwan. He is the former
president of Society of Ultrasound in Medicine, ROC (SUMROC), the former Chairman of Education
Committee of the Asian Federation of Societies for Ultrasound in Medicine and Biology (AFSUMB)
and the former Associate Editor of the Ultrasound in Medicine and Biology (the official journal of
World Federation of Ultrasound in Medicine and Biology, WFUMB). He is currently the President of
the Radiological Society of the ROC, Treasurer of the AFSUMB, Chairman of Education Committee of
the SUMROC, the Editor in Chief of the Journal of Medical Ultrasound (JMU, the official journal of the
AFSUMB).. Dr. Chou has been active in scientific and teaching programs nationally and internationally, and
has organized a number of international congresses, workshops, and symposia, particularly on ultrasound and
Imaging of the Breast and Emergency and Critical Care. He has expertise in all imaging modalities of abdominal
and breast diseases, and emergency medicine and critical care. His scientific interests include ultra-
sound image analysis, computer-aided diagnosis and detection, and ultrasound-guided minimally
invasive techniques.
Multimed Tools Appl
Chui-Mei Tiu is a chief and associate professor in the Child Radiology Division, Department of Radiology,
Taipei Veterans General Hospital and in National Yang Ming University, Taiwan, respectively. She also received
the MBA degrees from Ming Chuan University, Taiwan. Her research interests include: medical imaging
diagnosis, child radiology, breast imaging diagnosis, emergency radiology, and biomedical information.
Ying-Che Hsu was a graduate student in Department of Electrical Engineering at National Tsing-Hua University,
and received MS degree in 2009. His research interests include biomedical imaging processing.
Multimed Tools Appl
Din-Yuen Chan is a professor in the Department of Computer Science and Information Engineering, and a
director in the Department of Electrical Engineering at National Chiayi University, Taiwan. He received his MS
and PhD degrees from the Department of Electrical Engineering from National Cheng-Kung University, Taiwan,
in 1992 and 1996, respectively. His research interests include visual communication, visual-audio systems, image
processing, multimedia compression, and image watermarking.