Micron 61 (2014) 1–8

Contents lists available at ScienceDirect

Micron
journal homepage: www.elsevier.com/locate/micron

Accurate segmentation of leukocyte in blood cell images using

Atanassov’s intuitionistic fuzzy and interval Type II fuzzy set theory
Tamalika Chaira ∗
Centre for Biomedical Engineering, Indian Institute of Technology Delhi, New Delhi 110016, India

a r t i c l e i n f o a b s t r a c t

Article history: In this paper automatic leukocyte segmentation in pathological blood cell images is proposed using intu-
Received 13 December 2013 itionistic fuzzy and interval Type II fuzzy set theory. This is done to count different types of leukocytes
Received in revised form 17 January 2014 for disease detection. Also, the segmentation should be accurate so that the shape of the leukocytes is
Accepted 19 January 2014
preserved. So, intuitionistic fuzzy set and interval Type II fuzzy set that consider either more number of
Available online 12 February 2014
uncertainties or a different type of uncertainty as compared to fuzzy set theory are used in this work.
As the images are considered fuzzy due to imprecise gray levels, advanced fuzzy set theories may be
Keywords:
expected to give better result. A modified Cauchy distribution is used to find the membership function.
Cauchy distribution
Intuitionistic fuzzy set
In intuitionistic fuzzy method, non-membership values are obtained using Yager’s intuitionistic fuzzy
Fuzzy divergence generator. Optimal threshold is obtained by minimizing intuitionistic fuzzy divergence. In interval type
Type II fuzzy set II fuzzy set, a new membership function is generated that takes into account the two levels in Type II
Leukocyte segmentation fuzzy set using probabilistic T co norm. Optimal threshold is selected by minimizing a proposed Type II
fuzzy divergence. Though fuzzy techniques were applied earlier but these methods failed to threshold
multiple leukocytes in images. Experimental results show that both interval Type II fuzzy and intuition-
istic fuzzy methods perform better than the existing non-fuzzy/fuzzy methods but interval Type II fuzzy
thresholding method performs little bit better than intuitionistic fuzzy method. Segmented leukocytes
in the proposed interval Type II fuzzy method are observed to be distinct and clear.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction
Segmentation is an important step medical image processing
which is used for extracting different objects in an image. Espe-
cially for pathological images, segmentation is very much essential.
In pathological studies, blood cell parameters such as erythrocytes
(red blood cell), leukocytes (white blood cell), and platelets are
very essential to detect many diseases such as anemia, leukemia,
cancer and any other infections. Out of these blood parameters,
leukocyte plays an important role in human immune system. The
family of leukocyte is comprised of eosinophill, basophill, neu-
trophill, lymphocytes, monocytes. The five types of leukocytes can
be distinguished by their cytoplasmic granules, staining properties
of the granules, size of cell, the proportion of the nucleus to the
cytoplasmic material, and the type of nucleolar lobes.
Leukocyte counting and the type of leukocyte are essential in
order to detect diseases. Conventionally medical experts visually
count the cells but it is time consuming. Today, automated tech-
niques are carried out to perform medical image diagnosis. But still
∗ Tel.: +91 11 26596165.
E-mail address: tchaira@yahoo.com
accurate counting is difficult. So, for more accurate diagnosis, fuzzy
or intuitionistic fuzzy or Type-II fuzzy set are being explored (Fig. 1).
Many authors suggested different techniques for image segmen-
tation. But studies on pathological image segmentation are few.
Segmentation should be accurate in leukocyte segmentation as the
shape of the leukocytes should be preserved in order to detect the
disease depending on the number and type of leukocytes present in
the blood cell. Liao and Deng (2002) segmented white blood cells.
In this work, they thresholded the blood cells and then applied
morphology and finally shape detectors to find the shape. Piuri
and Scotti (2004) initially separated the leucocytes from the others
blood cells and then extracted morphological indexes and finally
classified the leucocytes. Scotti (2005) identified Leukemia using
morphological method. They first separated the leukocyte from
other blood cells and used morphological indexes and then clas-
sified the cells. Ghosh et al. (2010) suggested automatic leukemia
recognition using fuzzy divergence (Chaira and Ray, 2003a,b). They
proved that membership function from Cauchy distribution shows
better result. Textural approach to leukocyte recognition using
GLCM matrix was suggested by Sabino et al. (2004). Ramoser et al.
(2008) had given an idea of leukocyte segmentation and classifi-
cation in blood smear images. They used some features that are
related to cytoplasm and nucleus properties. The features used are
color, shape of the nucleus, and finally SVM classifier is used to

0968-4328/$ – see front matter © 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.micron.2014.01.004
2 T. Chaira / Micron 61 (2014) 1–8
Fig. 1. (a) Original image, (b) thresholded using Sugeno generator, (c) thresholded using Yager generator.
classify the images. Wu et al. (2008) segmented colored WBC
images using Otsu’s approach (Otsu, 1979) and HSI color model.
Adolla et al. (2008) provided a review on cell segmentation
techniques. Chan et al. (2010) proposed leukocyte nucleus seg-
mentation and nucleus lobe counting method. They extracted the
leukocyte region from blood smear images and developed a leuko-
cyte recognition system. The number of lobes increases when there
is leukemia, liver disease, cancer, vitamin B12 deficiency, etc.
But there is hardly any work using intuitionistic fuzzy set or
Type II fuzzy set theory. Intuitionistic fuzzy set theory considers
two uncertainties – membership and non-membership degree. As
selection of membership function is user defined and it varies from
person to person, so some kind of hesitation exists while defining
the membership function. Membership function may be Gaussian,
Gamma, triangular and so on. Due to this hesitation, hesitation
degree is considered and so the non membership degree in intu-
itionistic fuzzy set is not the complement of the membership degree
as in fuzzy set rather less than or equal to the complement of the
membership degree.
In Type II fuzzy set, the membership function in Type I (ordi-
nary) fuzzy set is considered as ‘fuzzy’. The membership function is
considered to lie in an interval range with upper and lower mem-
bership levels and so it may be termed as interval Type II fuzzy set.
In this paper, an automatic leukocyte segmentation method
from pathological blood cell image (abnormal and normal cell
images) using both intuitionistic fuzzy set and interval Type II
fuzzy set theory is proposed that preserves the shape if leukocytes
in order to count different types of leukocytes for different disease
detection. Modified Cauchy membership function from Cauchy
distribution is used to find the membership function of the image.
In case of intuitionistic fuzzy set, Yager’s intuitionistic fuzzy
generator is used to find the non membership function. Intu-
itionistic fuzzy divergence is used to find the optimum threshold
value. In case of interval Type II fuzzy set, probabilistic fuzzy T co
norm is used to form a new membership function using the two
membership levels of interval Type II fuzzy set. Proposed Type II
fuzzy divergence is used to find the optimal threshold. Experiment
is conducted on several cell images where it is observed that the
leukocytes are clearly detected.
The paper is organized as follows. Section 2 overviews the intro-
duction to Intuitionistic fuzzy set. Section 3 writes about Cauchy
distribution and the modified membership function derived from it.
Section 4 details the procedure of leukocyte segmentation. Section
5 writes about the introduction of Type II fuzzy set. Section 6 details
the procedure of segmentation using interval Type II fuzzy set. Sec-
tion 7 discusses and displays the results and finally the conclusion
is drawn in Section 8.
2. Introduction to Atanassov’s Intuitionistic fuzzy set
A fuzzy set A in a finite set X = {x1 , x2 , . . ., xn } may be represented
mathematically as:

A = {(x, A (x) x ∈ X}
where, the function A (x) : X → [0, 1] is measure of degree of
belongingness or membership function of an element x in the
finite set X and the measure of non-belongingness is 1 − A (x). An
intuitionistic fuzzy set A in a finite set X may be mathematically
represented as (Attanassov, 1999):
A = {(x, A (x), A (x))|x ∈ X} (1)
where, the functions A (x), A (x) : X → [0, 1] are respectively the
membership and non-membership functions of an element x in a
finite set X with the necessary condition
0 ≤ A (x) + A (x) ≤ 1
Stressing the necessity of taking into consideration a third param-
eter A (x), known as hesitation degree that arises due to the lack
of knowledge or the ‘personal error’ in assigning the membership
degree, an intuitionistic fuzzy set A in X may be represented as:

A = {(x, A (x), A (x), A (x)) x ∈ X}
with the condition
A (x) + A (x) + A (x) = 1 (2)
It is obvious that 0 ≤ A (x) ≤ 1, for each x∈X.
 T. Chaira / Micron 61 (2014) 1–8 3

3. Cauchy distribution is shown in Fig. 2 using Sugeno and Yager generator where it is
observed that the thresholded leukocytes either are not clear or
The Cauchy distribution has the probability density function contain noise with Sugeno type generator.
which is defined as: Using Yager’s generator, the non membership function is writ-
1 ten as:
f (x; a, ) = 2
(3)
(1 + (x − a/) ) A (gij ) = (1 − A (gij )˛ )
1/˛

where ‘a’ is the location parameter,  is a scale parameter.

For standard Cauchy distribution,  = 1 and a = 0, the distribution The hesitation degree is computed as:
becomes:
A (gij ) = 1 − A (gij ) − A (gij )
1
f (x; 0, 1) =
(1 + (x − a)2 ) Due to the hesitation, the membership lie in the interval
{A (gij ) + A (gij ), A (gij ) − A (gij )}.
To find the membership values of the image, the distribution is
The membership values of intuitionistic fuzzy set is modified as:
modified according
√ to the image requirement.
new (gij ) = A (gij ) + A (gij )
 = 1/ const, const = 1/(fmax − fmin ) where fmax , fmin are the A
Next fuzzy divergence is computed between the thresholded
maximum and minimum gray values of the image.
image and an ideally segmented image. Ideally segmented image is
Substituting the values, the distribution becomes:
√  such that the object pixels lie exactly in the object region and back-
√
const const ground pixels lie exactly in the background region, i.e. membership
f (x; a, ) = =
(1 + const · (x − a)2 )  values are all 1 (Chaira and Ray, 2003a,b).

×
1
,

M−1
N−1
new (gij )−B (gij )
(1 + const · (x − a)2 ) IFD(A, B) = 2 − (1 − new
A (gij ) + B (gij )) · e
A

√ j=0 i=0
where  = 1/ const
B (gij )−new (gij )
In order to make
√ the membership values feasible or practical, the − (1 − B (gij ) + new
A (gij )) · e
A

constant term ( const/) is not taken into account. So, the function
that is used to derive membership function is: Considering image B to be an ideally thresholded image, so substi-
tuting B (gij ) = 1, we get:
1
f (x, a) = (4)
(1 + const · (x − a)2 ) 
M−1
N−1
new (gij )−1
IFD(A, B) = 2 − (2 − new
A (gij )) · e
A
4. Leukocyte segmentation using Intuitionistic fuzzy
j=0 i=0
divergence
1−new (gij )
− new
A (gij ) · e
A (6)
Consider an image A of size M × N whose maximum and min-
imum values are noted. For a particular threshold, mean of the
background and foreground regions are computed. Then using Intuitionistic fuzzy divergence is computed for all the thresh-
Cauchy membership function, at a certain threshold ‘t’ that sepa- old gray levels. The threshold gray level at which the divergence is
rates the background and foreground region, the membership minimum is the optimum threshold.
function from Eq. (4) of the object (foreground) region is written For better segmentation, half of the optimum threshold value is
as: used to threshold the blood cell image to obtain segmented leuko-
1 cytes.
(aij ) = if, aij > t aij is the gray level
1 + const · (aij − mo )2
(5a) 5. Introduction to Type II fuzzy set

Type II fuzzy set (Mendel, 2007; Zadeh, 1975) is useful in situ-

If the image pixel gray level aij is more than the threshold ‘t’, the
ation where the membership values are such that it is difficult to
membership function will take the value in Eq. (5a).
agree with accurate membership values. This is due to the fact that
Likewise, the membership function A (aij ) of the background
some uncertainty may present in location or shape or other param-
region is
eters. Type II fuzzy set accounts this uncertainty by considering
1 another degree of freedom for better representation of uncertainty
(aij ) = if aij ≤ t (5b)
1 + const · (aij − mb )2 where the membership functions are themselves fuzzy. Type II
fuzzy set is obtained by blurring Type I fuzzy set. So, if the mem-
If the image pixel gray level aij is less than the threshold ‘t’, the bership function of Type I fuzzy set is blurred, Type II fuzzy set is
membership function will take the value in Eq. (5b).where m0 and obtained. So, one can define Type I fuzzy set and assign upper and
mb are the mean of object and background regions respectively lower membership degrees to each element to construct a Footprint
region and may be written as: of uncertainty (FOU). Fig. 2 shows the Type II fuzzy membership
t L−1 function (Ensafi and Tizoosh, 2005).
f · count(f )
f =0 f =t+1
f · count(f )
m0 =  , mb = L−1 Fuzzy sets of type II are the fuzzy sets for which the member-
t
f =0
count(f ) f =t+1
count(f ) ship function does not have a single value for every element but an
interval. A type-2 fuzzy may be written as:
To find the non membership function, intuitionistic fuzzy gen-

erators are used. Though there are many intuitionistic fuzzy ATypeII = {x, A (x)|x ∈ X}
generators such as Sugeno, Yaher’s generator in literature but in

this paper Yager’s generator (Yager, 1980) is used. A sample result where A (x) type-2 membership function.
4 T. Chaira / Micron 61 (2014) 1–8

Fig. 2. (a) Type I membership function, (b) Interval of Type II fuzzy membership function (shaded region) is the FOU.

When the FOU is uniform, it represents an interval Type II fuzzy In this experiment ˛ = 0.5 is chosen. On increasing or decreasing the
set and this can be described in terms of lower and upper member- value of ˛, no changes are observed. The limit of ˛ is 0 < ˛ < 1. It can
ship function. These are written as: take any value. A sample result is shown for different values of ˛ in
Fig. 3 where it is observed that no change is observed on changing
lower = [(x)]1/˛
(7) the values of ˛.
upper = [(x)]˛ Using Probabilistic sum, a new membership function is gener-
ated which is written as:
where ˛ ∈ (0, 1). So, a more practical form of representing Type II
fuzzy set is:
PA (gij ) = upper (gij ) + lower (gij ) − upper (gij ) · lower (gij ) (8)
ATypeII = {x, U (x), L (x)|x ∈ X}
So for each threshold gray level, mean of the object and background
and L (x) < (x) < U (x),  ∈ [0, 1].
regions are computed and the membership values of the pixels
are computed using modified Cauchy distribution. Then new mem-
6. Segmentation using interval Type II fuzzy set
bership function is computed using the two membership levels in
interval Type II fuzzy set as mentioned above.
In segmentation problem, T co norm is used to form a new
Fuzzy divergence is computed between the thresholded image
membership function using the two membership functions in inter-
and ideally segmented image. Ideally segmented image is such
val Type II fuzzy set. Before discussing the segmentation, a brief
that the membership values are all 1 (Chaira and Ray, 2003a,b) as
overview of fuzzy T co norm is given.
described earlier.
A T conorm S : [0, 1] → [0, 1] is a kind of binary operation used in
the framework of fuzzy logic and probabilistic metric space. It rep-
resents union in fuzzy set theory. It is an associative, commutative 
M−1
N−1
P (gij )−B (gij )
D(A, B) = 2 − (1 − PA (gij ) + B gij )) · e A
and increasing function such that S(0, x) = x is an associative for all
j=0 i=0
x ∈ [0, 1]. The four basic T co norms are:
(i) Maximum T co norm: SM (x,y) = max(x,y)
B (gij )−P (gij )
(ii) Probabilistic sum: SP (x,y) = x + y − x·y) − (1 − B (gij ) + PA (gij )) · e A

(iii) Lukasiewics T co norm: SL (x,y) = min(x + y,1)

(iv) Nilpotent T co norm: Sn (x,y) = max(x,y) if x + y < 1
Considering image B to be an ideally thresholded image and sub-
= 0, otherwise
Initially the image is fuzzified as per Eq. (5). Then upper and stituting B (gij ) = 1, we get:
lower membership functions of Type I fuzzy set are computed as:
 1/˛  
M−1 N−1
P (gij )−1 1−P (gij )
lower (gij ) =  (gij ) D(A, B) = 2 − (2 − PA (gij )) · e A − PA (gij )) · e A

 ˛ ˛ ∈ (0, 1)
j=0 i=0
upper (gij ) =  (gij )
(9)
with
1
 (g) = if aij > 1 for object Fuzzy divergence is computed for all the threshold gray levels.
(1 + const · (x − m0 )2 )
The threshold gray level at which the fuzzy divergence is minimum
1
= if aij ≤ t for background is the optimum threshold. For better segmentation, threshold is
(1 + const · (x − m1 )2 ) chosen as half of the optimum threshold.

Fig. 3. (a) original image, (b) thresholded with ˛ = 0.1, (c) thresholded with ˛ = 0.5, (d) thresholded with ˛ = 0.9.
 T. Chaira / Micron 61 (2014) 1–8 5

Fig. 4. Single leukocyte cell images: (a) thresholded image using Otsu’s method, (b) thresholded image using fuzzy set, (c) thresholded image intuitionistic fuzzy set, (d)
thresholded image using Type II fuzzy set.
6 T. Chaira / Micron 61 (2014) 1–8

Fig. 5. Abnormal leukocyte cell images: (a) thresholded image using Otsu’s method, (b) thresholded image using fuzzy set, (c) thresholded image using intuitionistic fuzzy
set, (d) thresholded image using interval Type II fuzzy set.
 T. Chaira / Micron 61 (2014) 1–8 7

Table 1
Threshold values of different images using different methods.

Images Otsu’s method Fuzzy method Int. fuzzy method Interval Type II fuzzy

Image1 196 244 196 156

Image2 197 243 196 178
Image3 192 246 191 154
Image4 199 245 198 182
Image5 197 242 197 167
Image6 198 241 197 161
Image7 199 243 202 158
M Image1 183 250 181 166
M Image2 189 252 184 200
M Image3 185 248 181 170
M Image4 192 250 193 171
M Image5 200 254 215 206
M Image6 220 250 223 205
M Image7 223 245 217 212

Algorithm: method where less noise is present around the cells. In the pro-
1. Select an image posed two methods – using Intuitionistic fuzzy and interval Type
2. For any particular threshold gray level, compute the mean of
the object and background regions.
II fuzzy methods, it is observed that both the methods give almost
3. Calculate the membership function of the pixels of the object similar and better results almost without noise than Otsu’s method
and background region from the mean of the and fuzzy method except Image 2, Image 3 and Image 7 where very
object/background region respectively using Cauchy little noise in intuitionistic fuzzy method. So, interval Type II fuzzy
membership function.
method gives much clearer and better segmented image.
4. Calculate intuitionistic/interval Type II fuzzy divergence using
Eq. (6) or Eq. (9). Fig. 5 shows the result on abnormal leukocyte images where
5. Calculate the divergence for all the threshold gray levels. 3 or 4 leukocytes are jumbled together which may be mature or
6. Select the minimum divergence value and mark the immature and it can be judged only by shape. Almost in all the
corresponding threshold gray level. images, fuzzy method shows unwanted noise and the outer shapes
7. Select half of the threshold value.
are not clear and especially in image such as Ab Image1, Ab Image2,
8. Threshold the image with the new threshold value.
Image 3, Ab Image4, Ab Image5, noise is so much that the cells are
not clearly differentiated. In other images there are lots of noise
7. Results and discussion
in the thresholded images around the leukocytes. Otsu’s method
shows little bit better than the fuzzy method. The thresholded
Experiments are preformed on 100 pathological images and
leukocytes contain small holes or white spots in them in Ab image
these are normal cell and abnormal cell (leukemia). The patho-
1, Ab image 3, Ab image 5, Ab image 7 and so the thresholded
logical images are obtained from medical science and technology
leukocytes are not clear and uniform. The proposed intuitionistic
laboratory in IIT Kharagpur. The subjects (blood samples) are
fuzzy method and Type II fuzzy method does not produce noise in
obtained from Midnapore medical college and hospital, Midnapore.
the segmented leukocytes, no holes or bright spots in the leuko-
Blood smear is prepared in a disinfected slide. After smear prepa-
cytes and the leukocytes are very prominent and clear. Among the
ration, the images are optically grabbed by Leica Observer (Leica
two proposed methods, intuitionistic fuzzy method contains some
DM750, Leica Microsystems) under 100 × oil objective (N.A. 1.515)
holes and bright spots in images Ab Image 1 and Ab Image 3 but
having 0.064 m resolution. The input images are the micropho-
no holes/spots are present in interval Type II fuzzy method. Inter-
tographs from blood smears.
val Type II fuzzy method produces clear and prominent leukocytes
Nucleus segmentation is carried out and so the shape of the
without any noise. So, interval Type II fuzzy method gives better
nucleus is considered. In one experiment, images containing nor-
result than intuitionistic fuzzy method.
mal cell and in other experiment, leukemic cell images are used
So, the shape obtained using the proposed interval Type II fuzzy
that contain mature and immature cells. In leukemia, there is an
method is almost accurate and is useful in counting different types
increase of immature/abnormal white blood cells. Segmentation
of leukocytes accurately. This will help the physicians to detect the
becomes difficult when many leukocytes (mature or immature)
disease accurately.
are present in the image and cells and are jumbled together. To
Table 1 shows the threshold of the images using different seg-
know the type of leukocytes, thresholding algorithm must be effi-
mentation methods. It is observed that the threshold values of
cient. In this paper results on 7 images containing one leukocyte
Otsu’s method and intuitionistic fuzzy method are almost closer
and 7 abnormal images containing more than one leukocyte are
as can be observed from the results. Threshold level of the fuzzy
shown. The results are compared with Otsu’s method (Otsu, 1979)
method is more and the images are not segmented properly.
and fuzzy method (Ghosh et al., 2010).
Threshold level using interval Type II fuzzy method is less and so
It is to be noted that if the image is thresholded using threshold
the images are segmented almost clearly.
value obtained in Otsu’s method, leukocytes will not be segmented.
So, half the threshold value obtained from Otsu’s method is used
to threshold the image. In the result section, half of the threshold 8. Conclusion
value is used.
Fig. 4 shows the result on cell images where one WBC is present. In this paper, an automatic leukocyte segmentation scheme on
Both intuitionistic fuzzy set and interval Type II set are used in pathological images (both abnormal and normal cells) using both
segmentation. Results on 7 images are shown and the results are intuitionistic fuzzy set and interval Type II fuzzy set theory is sug-
compared with Otsu’s method and fuzzy method. It is observed gested that preserves the shape of the leukocytes. This is done to
that the resulting fuzzy segmented image contains noise which is count different types of leukocytes in the blood cell for different
present around the cells and the shape is also not clear. In Otsu’s disease detection. While using intuitionistic fuzzy set, a modi-
method, thresholded images are little bit better than the fuzzy fied Cauchy membership function is used to find the membership
8 T. Chaira / Micron 61 (2014) 1–8
function of the image and then Yager’s intuitionistic fuzzy genera-
tor is used to find the non membership function. Intuitionistic fuzzy
divergence is used to find the optimum threshold value. In interval
Type II fuzzy set, probabilistic fuzzy T co norm is used to form a new
membership function using the two membership levels of interval
Type II fuzzy set. Type II fuzzy divergence is used to find the optimal
threshold. Experiment is conducted on several pathological images.
On comparing with Otsu’s method and fuzzy method for leuko-
cyte segmentation, it is observed that all types of leukocytes are
detected clearly and almost without any noise with the proposed
methods. But among intuitionistic fuzzy and interval Type II fuzzy
methods, interval Type II fuzzy method produces much clear and
prominent leukocytes with proper shape. Thus counting different
types of leukocytes present in blood cells can be accurately done
using the proposed interval Type II fuzzy method. So, interval Type
II fuzzy performs the best. The reason for better segmentation is
that Type II fuzzy set considers uncertainty in the fuzzy member-
ship function in Type I fuzzy set and so the membership function is
considered to lie in an interval range. There are two levels–upper
and lower levels of the membership function in interval Type II
fuzzy set. Then using t co norms, a new membership function is
constructed. So, the membership function used in Type I fuzzy set
is different from Type II fuzzy set.
Acknowledgements
Author would like to acknowledge Dr. Madhumala Ghosh, IIT
Kharagpur for providing the pathological images. The author would
also like to acknowledge the anonymous reviewers for suggesting
valuable comments in improving the quality of the manuscript.
Finally, the author would like to acknowledge the Department
of Biotechnology, Ministry of Science and Technology, Govt. of India
for financially supporting the research work carried out under the
National award scheme ‘Innovative Young Biotechnologist Award,
IYBA-2010”.
References
Adolla, R., et al., 2008. Blood cell image segmentation: a review. In: Abu Osman, N.A.,
Ibrahim, F., Wan Abas, W.A.B., Abd Rahmanm, H.S., Ting, H.N. (Eds.), Biomed
2008, 21. , pp. 141–144.
Attanassov, K.T., 1999. Intuitionistic fuzzy sets: Theory and Applications, Series in
Fuzziness and Soft Computing. Phisica-Verlag, Heidelberg.
Chaira, T., Ray, A.K., 2003a. Segmentation using fuzzy divergence. Pattern Recogni-
tion Letters 4 (12), 1837–1844.
Chaira, T., Ray, A.K., 2003b. Threshold selection using fuzzy set theory. Pattern Recog-
nition Letters 5 (8), 865–874.
Chan, Y.-K., et al., 2010. Leukocyte nucleus segmentation and nucleus lobe counting.
BMC Bioinformatics 11, 558.
Ensafi, P., Tizoosh, H.R., 2005. Type II fuzzy image enhancement. In: Kamel, M.,
Campilho, A. (Eds.), Lecture Notes in Computer Sciences 3656. Springer-Verlag,
Berlin Heidelberg, pp. 159–166.
Ghosh, M., et al., 2010. Automatic leukocyte recognition using fuzzy divergence.
Micron 41, 840–846.
Liao, Q., Deng, Y., 2002. An accurate segmentation method for white blood cell
images. In: Proc. of IEEE International Symposium on Biomedical Imaging.
Mendel, J.M., 2007. Type 2 Fuzzy sets and systems: an overview. IEEE Computational
Intelligence Magazine 2 (1), 20–29.
Otsu, N., 1979. A threshold selection method from gray level histograms. IEEE Trans-
action on System, Man, and Cybernetics 9, 62–66.
Piuri, V., Scotti, F., 2004. Morphological classification of blood leucocytes by micro-
scope images. In: Proc. International Symposium on Computational Intelligence
for Measurement Systems and Applications, Boston, MD, USA.
Ramoser, H., Laurain, V., Bischof, H., Ecker, R., 2008. Leukocyte segmentation and
SVM classification in blood-smear images. International Journal of Machine
Graphics & Vision 17 (1), 187–200.
Sabino, D.M.U., et al., 2004. A texture approach to leukocyte recognition. Real-Time
Imaging 10 (4), 205–216.
Scotti, F., 2005. Automatic morphological analysis for acute leukemia identifica-
tion in peripheral blood microscope images. In: Proc of IEEE International
Conference on Computational Intelligence for Measurement Systems and
Applications.
Wu, J., et al., 2008. A novel color image segmentation method and its application to
white blood cell image analysis. In: Proc. of IEEE-International Conference on
Signal Processing, vol. 2, pp. 245–248.
Yager, R.R., 1980. On the measures of fuzziness and negation. Part II: lattices. Infor-
mation and Control 44, 236–260.
Zadeh, L.A., 1975. The concept of a linguistic variable and its application to approx-
imate reasoning-I. Information Sciences 8, 199–249.