Available online at www.sciencedirect.

com

ScienceDirect
Procedia Computer Science 46 (2015) 1449 – 1457

International Conference on Information and Communication Technologies (ICICT 2014)

RGB Histogram based Color Image Segmentation Using

Firefly Algorithm
V. Rajinikantha,*, M. S. Couceirob
a
Department of EIE, St. Joseph’s College of Engineering,Chennai 600 119, Tamil Nadu, India.
b
Institute of Systems and Robotics, University of Coimbra, Ingeniarius, Lda.,Rua da Vacariça, n.37, 3050-381, Mealhada, Portugal.

Abstract

In this paper, optimal multi-level image segmentation is proposed using the Firefly Algorithm (FA). In this work, RGB
histogram of the image is considered for bi-level and multi-level segmentation. Optimal thresholds for each colour component
are attained by maximizing Otsu’s between-class variance function. The proposed segmentation procedure is demonstrated using
standard RGB dataset and validated using the existing FA in the literature combined with three randomization search strategies,
such as Brownian Distribution, Lévy Flight and the Gaussian distribution related random variable. The performance assessment
between FAs is carried out using parameters, such as objective value, PSNR, SSIM and CPU time.

© 2015
© 2014 The
The Authors.
Authors. Published
Published by
by Elsevier
ElsevierB.V.
B.V.This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of organizing committee of the International Conference on Information and Communication
Peer-review
Technologies under responsibility
(ICICT 2014). of organizing committee of the International Conference on Information and Communication
Technologies (ICICT 2014)
Keywords: RGB histogram; Segmentation; Otsu; Firefly algorithm; PSNR; SSIM; CPU time.

1. Introduction

Image segmentation is an essential procedure, being extensively considered to extract meaningful information
from grey scale or colour (RGB) images. During the segmentation process, a digital image is separated into multiple
regions, or objects, in order to extract and interpret the relevant information. In recent years, this procedure has been
widely considered in many key fields, such as remote sensing3,4,5 medical imaging16, and pattern recognition9.

* Corresponding author. Tel.: +91 9380593801.

E-mail address :[email protected]

1877-0509 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of organizing committee of the International Conference on Information and Communication Technologies (ICICT 2014)
doi:10.1016/j.procs.2015.02.064
1450 V. Rajinikanth and M.S. Couceiro / Procedia Computer Science 46 (2015) 1449 – 1457

Determining the exact threshold level to separate an image into desirable objects (foreground) from background
remains an extremely significant step in imaging science.

In the literature, a considerable number of parametric and nonparametric bi-level and multi-level thresholding
procedures have been proposed and implemented mainly for grey scale images1,2,4,11. Among them, global
thresholding is considered as the most preferred image segmentation technique because of its simplicity, robustness,
accuracy, and competence14. In general, existing parametric thresholding approaches are computationally costly,
time consuming, and some times the performance degrades depending on the image quality6,9. Nonparametric
traditional approaches, on the other hand, methods such as Otsu, Kapur, Tsai, and Kittler are simpler and successful
for bi-level thresholding11. When the number of threshold level increases, the complexity of thethresholding
problem also increases and the traditional method requires more computational time. Therefore, to overcome the
computational complexity of most traditional methods, heuristic based bi-level and multi-level image thresholding
procedures have been widely proposed by researchers for grey scale1,2, RGB10, multi-spectral and hyperspectral
images3,5. Recent meta-heuristic algorithms, such as cuckoo search1, bee colony2, and firefly14, are also employed to
solve the m-level image thresholding problem. Most of the above discussed methods are applied and validated on a
class of grey scaled images.
In recent years, the segmentation of RGB images, or more generally multi-spectral images, is also getting the
attention of researchers. The authors from Ghamisi et al. proposed a heuristic-based segmentation technique for a
class of hyperspectral colour images3,5. Su and Hu discussed a colour image quantization technique using self-
adaptive differential evolution algorithm and the technique was validated using standard test images10. Sarkar and
Das proposed a colour image segmentation procedure using Tsallis entropy and differential evolution. The authors
validated the proposed method using a class of RGB images using 2D histogram technique 12.
In the proposed work, the RGB histogram of the colour image is considered to solve the m-level thresholding
problem. The maximization of Otsu’s between-class variance function is chosen as the objective function. The
proposed segmentation procedure is a nonparametric approach, thus employing heuristic methods, such as Brownian
search based Firefly Algorithm (BFA), Lévy Flight based Firefly Algorithm (LFA) and FA with Gaussian
distribution related random variable (ε). The proposed method is implemented and validated on standard colour
images.

2. Problem formulation

Otsu’s based image thresholding was initially proposed back in 19798. This method returns the optimal threshold of
a given image by maximizing the between-class variance function. This procedure already proved its efficiency on
grey scale 2,4,7,11,14 and colour images 3,5.
In this paper, Otsu’s approach is considered for colour image segmentation with the aid of the RGB histogram. In
RGB space, each colour pixel of the image is a mixture of Red, Green, and Blue (RGB) and for that same image, the
data space size is [0, L-1]3 (R = [0, L-1], G = [0, L-1], and B = [0, L-1]). In spite of this, one can formalize the
heuristic based segmentation procedure as it follows5.
For a given RGB image, let there be L intensity levels in the range [0,1,2,…, L-1]. Then, the probability distribution
PiC can be defined as:

L1
hic
piC
N ¦ piC 1 (1)
i 0

where i is a specific intensity level in the range { 0 d i d L  1 } for the colour component C = {R,G,B}, Nis the total
number of pixels in the image, and hiC is the number of pixels for the corresponding intensity level I in component C.

L1
The total mean of each component of the image is calculated as: PTC ¦ ipiC 1 (2)
i 0
 V. Rajinikanth and M.S. Couceiro / Procedia Computer Science 46 (2015) 1449 – 1457 1451

Them-level thresholding presents m-1 threshold levels t cj , where j = 1,2,…,m-1, and the operation is performed as:

­
°0 ,
°1 f C ( x , y ) d t1C
° ( t C  t C ),
°2 1 2 t1C  f C ( x , y ) d t2C
°
F C ( x , y ) ®  (3)
° C  f C ( x, y ) d tC
° 1 ( t C  t C ), tm 2 m1
° 2 m2 m1
° f C ( x , y ) ! tmC
1
°L  1,
¯

wherein x and y are the width (W) and height (H), in pixels, of the image of size H × W denoted by f C (x, y) with L
intensity levels for each component.
c are given by:
The probabilities of occurrence wCj of classes Dic ,…, Dm

­ tC C
°
° ¦i j 0 pi , j 1
°
°° tC
wC
j ®
°
¦ j
i tC
j 1
 1 piC , 1 j  m (4)
°
L1
°
°¦
°¯
i tC
j 1
 1 piC , j m

The mean of each class P Cj can then be calculated as:

­ C
tC p
°
°
°
¦i j 0 wCi q , j 1
j
°
tC pC
: P Cj °
®
°
¦ j
i tC
1 i ,
j 1 w j
C
1 j  m (5)
°
° L1 pC j m
°
°
¦i tCj 1 1 wiC ,
¯ j

At last, Otsu’s between-class variance of each component can be defined as:

2 m 2
c
VB ¦ wCj §¨© PCj  PTC ·¸¹ (6)
j 1

where wCj is the probability of occurrence. The m-level thresholding is reduced to an optimization problem to search

for t Cj , that maximizes the objective function (Jmax) of each image component C being defined as:

MC max c2 ( t C )
VB for C ={R,G,B} (7)
C j
1ti ,,L1

Solving this optimization problem for an RGB image may require a much larger computational effort for both bi-
level and multi-level thresholds. Many methods have been proposed in the literature to solve the image thresholding
problem6,9,13. Compared to traditional analytical techniques, heuristic-based segmentation techniques are used as
alternatives due to their computational efficiency. Next section briefly describes some of these.
1452 V. Rajinikanth and M.S. Couceiro / Procedia Computer Science 46 (2015) 1449 – 1457

3. Brief overview of algorithms in the study

In this paper, the Firefly Algorithm (FA) and its recent improved forms are considered. The classical FA was
initially proposed by Yang19. It is a nature-inspired meta-heuristic algorithm, in which flashing illumination patterns
generated by invertebrates, such as glowworms and fireflies, were at the essence of its creation15.
The traditional FA is developed by considering the following conditions17,18,20:
(i) Fireflies are unisex and one firefly will be attracted towards the nearest firefly regardless of its sex;
(ii) The attractiveness between two fireflies is proportional to the luminance;
(iii) The brightness of a firefly is somehow related with the analytical form of the fitness or cost function
assigned to guide the search process. For instance, in a maximization problem, the luminance of a firefly
is considered as to be directly proportional to the value of cost function (i.e., the luminance is the fitness
function).
The movement of the attracted firefly i towards a brighter firefly j can be determined by the following position
update equation:
γ d 2
ij
X it 1 X it  β0 e ( X tj  X it )  randomization parameter (8)

γ d 2
ij t
where X it 1 is the updated position of firefly, X it is the initial position of firefly, and β0e ( X j  X it ) may be
considered as the attractive force between fireflies.
The parameterizations of the algorithm, namely the necessary parameters to update the position of a firefly, have
been discussed in the literature. In a recent paper from Raja et al.15, the following three random parameters, such as
Brownian search based FA (eq. 9), Lévy flight based FA (eq. 10), and the traditional FA, were considered to update
the position of fireflies.
D1. sign (rand - 1 2 ) † B(s) (9)
D1. sign (rand - 1 2 ) † L(s) (10)
D1. Ni (0,1) (11)

where L( s ) A. s
1/ E , B( s ) A. s
D / 2 , A E* ( E ) sin§¨ ES ·¸ 1 . A is a random variable, β is the spatial exponent, α is the
© 2 ¹S
temporal exponent, and * ( E ) is the Gamma function.
Initial firefly algorithm parameters are assigned based on the discussion presented by Raja et al.15 which is
summarized in Table 1.

Table.1Initial parameters of heuristic algorithms

Parameter Values
Number of Iterations 250
Population 20
Search dimension m
Stopping criteria Jmax

4. Implementation

The grey level thresholding problem deals with finding the most favourable thresholds within the range [0, L-1] that
maximize a fitness criterion. Similarly, considering the RGB histogram based technique, the heuristic algorithm
finds the optimal thresholds within the data space of [0, L-1]3 by maximizing Otsu’s between-class variance
function. The dimension of the segmentation problem mainly depends on the required threshold (m) levels. In this
work, for the colour image segmentation problem, heuristic algorithms are allowed to explore [[0, L-1]3] m data
space in order to obtain the optimal threshold levels. Hence, RGB histogram based colour image segmentation is a
 V. Rajinikanth and M.S. Couceiro / Procedia Computer Science 46 (2015) 1449 – 1457 1453

challenging work when compared to its grey level alternative.

The quality of the segmented image is assessed using well-known image metrics, such as the Peak Signal-to-Noise
Ratio (PSNR) and Structural Similarity Index Matrix (SSIM). Additionally, both fitness function value Jmax and CPU
time are considered.
The PSNR gives the similarity of the segmented image against the original image based on the Mean Square Error
(MSE) of each pixel11,14:
§ 255 ·
PSNR(o,s) 20 log 10 ¨ ¸ ; dB (12)
¨ MSE( o, s ) ¸
© ¹
H W
RMSE( o,s ) MSE( x , y )
1
MN ¦¦[ o( i, j )  s( i, j )] 2 (13)
i 1j 1
where o and s are the original and segmented images of size H x W.
The SSIM is generally used to estimate the image superiority and inter-dependencies between the original and the
processed image2.
( 2Po Ps  C1 )( 2V os  C2 )
SSIM(o,s) (14)
( Po2  Ps 2  C1 )( V 2  V 2  C2 )
o s
where μo and μs are the average of o and s, σo2and σs2are the variance of o ands, σos is the covariance of o and s, and
C1= (k1L)2 and C2= (k1L)2 stabilize the division with weak denominator, with L = 256, k1 = 0.01, and k2 = 0.03.

5. Experimental results and discussions

The RGB histogram based image segmentation experiment is implemented in Matlab R2010a on an Intel Dual
Core 1.6 GHz CPU, 1.5GB RAM running window XP. The implemented segmentation procedures are a revised
form of the segmentation technique given at Matlab central webpage†. The proposed method is tested on standard
RGB test images (481 X 321sized), such as Butterfly, Star fish, Rhino, Horse, Flower, and Train‡. The number of
thresholds (m) considered in this procedure are 2, 3, 4 and 5. For each image, and for each m, the segmentation
procedure is repeated 15 times and the mean value of the trials is chosen as the set of optimal thresholds and
performance measures.
Initially the BFA, LFA, and conventional FA based optimization procedure is tested on the Butterfly image for m =
2-5. Fig. 1 (a - f) shows the original image, RGB histogram, segmented image and the corresponding optimal RGB
threshold values. From Fig.1 (c - f), one can observe that, the RGB image segmentation is a much more complicated
problem due to the three different colour patters, namely the Red (R), Green (G) and Blue (B) components. As
previously stated, the histogram of a RGB image is more complex when compared to the histogram of grey scale
image. Finding an optimal threshold on such complex histogram may be a challenging task. In other words, each
colour distribution should be separately analysed considering the RGB histogram, which may increase the
computational time. Fig. 2 shows the convergence of firefly algorithm based for m = 5. From this it is noted that all
the algorithms provide approximately similar performance. From Table 2 and Fig. 2 one can observe that the
convergence of LFA is better when compared with the alternatives considered in this study.
The above said procedure is repeated for other test images shown in Table 3. This table shows original 481 × 321
sized colour images, RGB histogram, and segmented bi-level and multi-level images with Brownian search FA
(BFA). The performance measure values for these images, such as objective function, PSNR, SSIM, and the CPU
time are presented in Table 2. The corresponding optimal thresholds (R, G, B) are presented in Table 4.

†
http://www.mathworks.com/matlabcentral/fileexchange/authors/117313
‡
http://www.eecs.berkeley.edu/Research/Projects/CS/vision/bsds/BSDS300/html/dataset/images.html
1454 V. Rajinikanth and M.S. Couceiro / Procedia Computer Science 46 (2015) 1449 – 1457

250

200

Pixel levels
150

100

50

0
0 50 100 150 200 250
RGB levels
(a) Original image (b) RGB histogram

300 300 300 300

200 200 200 200

100 100 100 100

0 0 0 0
0 50 100 150 0 50 100 150 200 0 50 100 150 200 0 50 100 150 200

(c) m=2 (d) m = 3 (e) m = 4 (f) m = 5

Fig.1. Segmentation of Butterfly image with BFA algorithm for m = 2-5

Table 2. Comparison of performance measure values for the RGB test images (mean value of 15 trials)
Objective function PSNR (dB) SSIM CPU time (min)
Image m
BFA LFA CFA BFA LFA CFA BFA LFA CFA BFA LFA CFA
2 3515.92 3402.61 3617.38 10.866 10.243 11.026 0.6399 0.6402 0.6394 0.2281 0.2826 0.2012
Butterfly

3 3629.37 3638.81 3640.72 14.297 15.173 14.927 0.7133 0.6936 0.7047 0.4032 0.4529 0.4173
4 3691.66 3669.02 3690.81 17.562 17.283 17.602 0.7835 0.7669 0.7639 0.4967 0.4838 0.4770
5 3822.81 3792.55 3811.01 19.554 19.328 19.715 0.8472 0.8317 0.8274 0.5503 0.5138 0.5259
2 1986.97 1972.10 1985.11 11.513 13.272 12.267 0.7320 0.7461 0.7392 0.3899 0.2139 0.2011
Star fish

3 2017.18 2081.66 2088.41 14.868 14.792 14.901 0.7831 0.7706 0.7593 0.3901 0.3103 0.3348
4 2107.25 2109.91 2098.77 18.382 18.281 18.332 0.8032 0.7996 0.8106 0.4825 0.3628 0.3915
5 2251.73 2178.24 2201.62 19.191 20.037 20.097 0.8529 0.8274 0.8461 0.5765 0.5100 0.47729
2 2004.99 2081.84 2107.28 9.881 11.368 11.206 0.6837 0.7106 0.7083 0.2925 0.26398 0.2337
Rhino

3 2216.72 2205.22 2192.77 13.463 14.122 13.974 0.7153 0.7342 0.7311 0.4107 0.4092 0.4099
4 2251.33 2267.18 2222.90 16.182 16.001 16.189 0.7316 0.7628 0.7528 0.4829 0.4415 0.43978
5 2388.16 2371.97 2382.28 18.068 17.926 17.874 0.7829 0.8152 0.7902 0.5719 0.5081 0.4866
2 2635.11 2671.03 2587.99 10.517 12.015 11.739 0.6402 0.7261 0.7264 0.2688 0.2179 0.2510
Horse

3 2688.04 2683.31 2660.37 14.701 14.826 14.519 0.7026 0.7418 0.7302 0.3519 0.28934 0.3218
4 2717.37 2716.03 2700.83 16.576 16.478 17.005 0.7792 0.7902 0.7886 0.4820 0.3826 0.3775
5 2782.70 2763.44 2746.67 18.269 20.027 20.157 0.8218 0.8142 0.8213 0.5792 0.4811 0.5337
2 1159.57 1302.61 1288.92 11.284 13.721 12.826 0.6820 0.7227 0.7301 0.3017 0.3122 0.3108
Flower

3 1420.23 1472.71 1392.44 16.168 14.916 16.026 0.7211 0.7529 0.7329 0.3725 0.3518 0.3597
4 1681.16 1592.88 1562.39 21.174 20.177 20.291 0.7938 0.8111 0.8102 0.4826 0.4114 0.4092
5 1690.00 1623.71 1607.35 20.844 21.002 20.926 0.8315 0.8268 0.8331 0.5639 0.5297 0.5442
2 1829.01 1803.55 1831.63 12.648 12.579 12.739 0.6826 0.6901 0.6883 0.3721 0.3301 0.3100
3 1873.77 1894.00 14.282 14.138 0.6869 0.7039 0.4028 0.3891
Train

1903.28 14.620 0.7132 0.3877

4 1937.42 1903.18 1917.22 18.548 18.207 18.442 0.7385 0.7835 0.7893 0.4927 0.4110 0.4072
5 1975.56 1955.28 1977.61 20.031 20.379 20.715 0.8193 0.8352 0.8374 0.5783 0.5117 0.4871
 V. Rajinikanth and M.S. Couceiro / Procedia Computer Science 46 (2015) 1449 – 1457 1455

Table 3. Test images, RGB histogram, and segmented images

Segmented images
Name Original Image Histogram m=2 m=3
m=4 m=5
250

200

Number of pixels
Star fish

150

100

50

0
0 50 100 150 200 250
RGB levels
120

100

Number of pixels
80
Rhino

60

40

20

0
0 50 100 150 200 250
RGB levels
200

150
Number of pixels
Horse

100

50

0
0 50 100 150 200 250
RGB levels
300

250
Number of pixels

200
Flower

150

100

50

0
0 50 100 150 200 250
RGB levels
200

150
Number of pixels
Train

100

50

0
0 50 100 150 200 250
RGB levels

4200

3600
Objective Function

3000

2400

1800
BFA LFA CFA
1200

600

0
40 80 120 160 200
Iteration number
Fig. 2. Convergence of FA search
1456 V. Rajinikanth and M.S. Couceiro / Procedia Computer Science 46 (2015) 1449 – 1457

Table 4. Optimal threshold values obtained for the RGB images with firefly algorithms
BFA LFA CFA
Image m
R G B R G B R G B
2 14,98 7,107 4,136 16,102 8,115 4,143 16,100 9,105 6,138
3 13,69,144 6,96,174 3,82,167 15,71,149 7,93,171 3,85,164 15,68,147 8,98,177 4,80,168
Butterfly

4 12,51,105, 6,71,124, 3,64,108, 13,49,107, 5,74,126, 3,61,110, 13,50,104, 6,72,123, 3,61,111,

146 178 172 145 181 175 144 179 175
12,54,96, 5,52,114, 2,46,107, 11,56,99, 4,50,115, 2,41,109, 11,57,95, 4,54,116, 2,44,109,
5
133,167 153,192 140,179 136,171 151,195 146,184 135,168 151,191 148,184
2 18,86 13,115 7,147 17,91 14,103 8,142 19,77 13,119 11,152
3 16,78,124 10,89,162 7,69,175 15,81,127 12,75,160 7,61,181 17,60,127 11,81,167 8,77,174
Star fish

15,71,122, 8,80,131, 5,51,94, 12,68,120, 10,65,137, 4,47,90, 14,58,129, 9,77,134, 6,50,91,

4
138 153 194 133 158 188 128 150 198
12,53,103, 5,33,107, 2, 49,92, 9,51,111, 7,30,101, 2, 45,84, 10,49,87, 7,31,112, 4, 51,91,
5
135,164 148,210 124,172 141,162 142,216 121,193 133,166 140,212 127,170
2 17,95 14,134 12,155 17,91 15,137 13,152 16, 99 13,131 10,153
3 15,61,137 13,79,174 10,43,158 14,66,134 13,82,171 11,41,165 15,58,132 11,71,167 9,47,10
12,43,111, 11,68,139, 8,35,98, 12,39,116, 11,71,134, 8,38,102, 14,41,104, 9,66,133, 8,31,104,
Rhino

4
152 186 204 147 182 196 147 180 201
10,44,97, 7,28,107, 5, 39,105, 10,49,92, 8,31,101, 4, 33,111, 12,48,93, 7,33,112, 6, 31,102,
5
145,177 148,202 166,221 147,170 142,193 174,208 140,172 141,196 161,218
2 11,118 7,112 2,123 14,113 8,118 3,132 10,120 7,115 4,1131
3 14,81,171 9,94,148 3,108,160 13,77,167 7,89,142 4,111,163 11,84,166 10,91,145 2,112,162
16,67,125, 10,81,120, 4,78,114,1 12,66,121, 8,86,122, 6,71,112, 9,62,128,1 8,84,125, 2,71,104,
Horse

4
181 157 70 180 164 173 78 163 176
18,54,95,1 12,71,112, 3,66,107,1 12,50,102, 10,76,117, 3,62,112, 13,44,99, 10,76,116 3,63,97,
5
38,196 130,171 42,184 145,193 135,178 144,187 131,186 , 133,178 152, 186
2 13,155 12,141 4,132 14,161 12,148 4,132 16,162 14,145 7,148
3 15,106,16 10,76,152 6,130,178 13,111,16 11,73,150 6,130,178 14,112,172 11,68,160 6,126,167
Flower

17,93,130, 9,85,127, 5,101,139, 11,73,128, 9,75,121, 5,101,139, 15,98,135, 10,82,128 5,95,127,

4
188 183 176 182 180 176 181 , 177 163
14,90,115, 9,46,88, 4,98,135,1 15,91,121, 9,38,83, 4,98,135,1 14,91,112, 9,38,68, 4,91,109,
5
154,198 124,189 61,177 151,186 128,184 61,177 151,190 119, 183 158,170
2 16,166 12,145 5,138 18,152 14,144 5,138 17,164 13,137 8,151
3 15,122,18 11,101,17 4,115,169 16,124,18 12,122,17 4,115,169 16,120,176 11,97,166 6,111,143
14,136,17 8,70,128, 5,102,148, 12,133,17 10,76,123, 5,102,148, 14,129,168 10,71,122 5,101,135,
Train

4
5,197 188 184 0,182 180 184 ,184 , 179 177
16,118,16 6,72,112, 4,85,127, 14,108,15 11,70,102, 4,85,127, 14,108,164 9,70,104, 4,65,97,
5
0,180,202 175,196 165,188 3,172,192 176,189 165,188 ,177,204 163,183 145,191

From these results, it is notable that despite small differences, all algorithms seem to reach the vicinities of the
optimal solution. For all the tested images with various threshold levels, the convergence time of both LFA and FA
seem better than BFA. On the other hand, the overall Jmax (objective function) values obtained with the BFA are
generally superior when compared to the alternatives.

6. Conclusions

In this paper, a new multi-level segmentation technique based on RGB histogram is proposed using Brownian
search based Firefly Algorithm (BFA), Lévy search based Firefly Algorithm (LFA), and conventional Firefly
Algorithm (FA). The proposed techniques are used to solve Otsu’s problem for delineating multilevel threshold
values. The segmentation procedure is validated using both qualitative and quantitative analysis, including
traditional measures, such as objective function, PSNR, SSIM, and CPU time, which are evaluated by converting the
segmented colour image into a grey scale image. Results demonstrate that the LFA and FA algorithms depict a
faster convergence when compared to BFA, while the latter is able to achieve a superior final objective function.
 V. Rajinikanth and M.S. Couceiro / Procedia Computer Science 46 (2015) 1449 – 1457 1457

References
1. Agrawal S, Panda R, Bhuyan S, Panigrahi BK. Tsallis entropy based optimal multilevel thresholding using cuckoo search algorithm,
Swarm and Evolutionary Computation, 2013; 11: 16–30.
2. Akay BA. Study on particle swarm optimization and artificial bee colony algorithms for multilevel thresholding. Applied Soft Computing,
2013; 13(6): 3066–3091.
3. Ghamisi P, Couceiro, MS, Benediktsson JA. Classification of hyperspectral images with binary fractional order Darwinian PSO and random
forests. SPIE Remote Sensing, 2013; 88920S-88920S-8.
4. Ghamisi P, Couceiro MS, Benediktsson JA, Ferreira N M F. An efficient method for segmentation of images based on fractional calculus
and natural selection. Expert Syst. Appl., 2012; 39 (16):12407– 12417.
5. Ghamisi P, Couceiro MS, Martins, FML, Benediktsson JA. Multilevel image segmentation based on fractional-order Darwinian particle
swarm optimization. IEEE T. on Geoscience and Remote sensing,2014; 52(5):2382-2394.
6. Lee SU, Chung SY, Park RHA. Comparative Performance Study Techniques for Segmentation, Computer Vision, Graphics and Image
Processing, 1990; 52 (2) : 171 - 190.
7. Manikantan K, Arun BV, Yaradonic DKS. Optimal Multilevel Thresholds based on Tsallis Entropy Method using Golden Ratio Particle
Swarm Optimization for Improved Image Segmentation, Procedia Engineering, 2012; 30 :364 – 371.
8. Otsu NA. Threshold selection method from Gray-Level Histograms, IEEE T. on Systems, Man and Cybernetics, 1979; 9 (1): 62-66.
9. Pal NR, Pal SK. A review on image segmentation techniques, Pattern Recognition, 1993; 26 (9): 1277 – 1294.
10. Su Q, Hu Z. Color Image Quantization Algorithm Based on Self-Adaptive Differential Evolution, Computational Intelligence and
Neuroscience, Vol. 2013, Article ID 231916, 8 pages.
11. Rajinikanth V, Sri Madhava Raja N, Latha K. Optimal Multilevel Image Thresholding: An Analysis with PSO and BFO Algorithms. Aust.
J. Basic and Appl. Sci., 2014; 8(9): 443-454.
12. Sarkar S, Das S. Multilevel image thresholding based on 2D histogram and maximum Tsallis entropy – A Differential Evolution Approach.
IEEE T. on Image Processing, 2013.; 22(12): 4788-4797.
13. Sezgin M, Sankar B. Survey over Image Thresholding Techniques and Quantitative Performance Evaluation, Journal of Electronic
Imaging, 2004; 13(1): 146 - 165.
14. Sri Madhava Raja, N.; Rajinikanth, V.; and Latha, K. Otsu Based Optimal Multilevel Image Thresholding Using Firefly
Algorithm, Modelling and Simulation in Engineering, vol. 2014, Article ID 794574, 17 pages.
15. Sri Madhava Raja N, Suresh Manic K, RajinikanthV. Firefly Algorithm with Various Randomization Parameters: An Analysis, In B.K.
Panigrahi et al. (Eds.): SEMCCO 2013, Part.1, Lecture notes in computer science (LNCS 8297), 2013;110–121.
16. Taher F, Werghi N, Al-Ahmad H, Donner C. Extraction and Segmentation of Sputum Cells for Lung Cancer Early Diagnosis. Algorithms,
2013; 6: 512-531.
17. Yang XS. Firefly algorithm, Lévy flights and global optimization, In Research and Development in Intelligent Systems XXVI, Springer-
Verlag London, UK, 2010; 209–218.
18. Yang XS. Firefly algorithm, stochastic test functions and design optimization, International Journal of Bio-inspired Computation, 2010;
2(2): 78- 84.
19. Yang XS. Firefly algorithms for multimodal optimization, Stochastic Algorithms, Foundations and Applications, SAGA 2009, Lecture
Notes in Computer Sciences 5792, 2009; 169–178.
20. Yang XS. Nature-Inspired Metaheuristic Algorithms, Luniver Press, Frome, UK, 2nd edition, 2011.