Escolar Documentos
Profissional Documentos
Cultura Documentos
ISBN 978-1-4799-3158-3
S.Karthick
Assistant.Professor,
Dept. of EEE,
The Kavery Engg. College,
Salem, India,
researchkarthick@gmail.com
The conventional hard clustering methods
classify each point of the data set just to single
cluster. As a consequence, the results are often very
crisp, i.e., in image clustering each pixel of the image
belongs just to one cluster. Still, in many real
situations, issues such as limited spatial resolution,
poor contrast, noise, overlapping intensities as well as
intensity inhomogeneities reduce the effectiveness of
hard (crisp) clustering methods. Fuzzy set theory [3]
has introduced the idea of partial membership,
described by a membership function. Fuzzy
clustering, as a soft segmentation technique, has been
widely studied and successfully applied in image
clustering and segmentation [4]-[9]. Among the fuzzy
clustering methods, fuzzy c-means (FCM) algorithm
is the most popular method used in image
segmentation because it has robust characteristics for
ambiguity and can retain much more information
than hard segmentation methods [11].
Even though the conventional FCM
algorithm works well on most noise-free images, it is
very sensitive to noise and other imaging artifacts,
since it does not consider any information about
spatial context. To give back this drawback of FCM,
a preprocessing image smoothing step has been
proposed [9], [12] and [13]. However, through
smoothing filters important image details can be lost,
especially boundaries or edges. Moreover, there is no
way to control the trade-off between smoothing and
clustering. Thus, a lot of researchers have
incorporated local spatial information into the
original FCM algorithm to improve the performance
of image segmentation. Ahmed et al.[9] proposed
FCM_S where the objective function of the classical
FCM is modified in order to give back the intensity
inhomogeneity in addition to allow the labeling of a
pixel to be influenced by the labels in its immediate
neighborhood. One disadvantage of FCM_S is that
the neighborhood labeling is computed in each
iteration step, something that is incredibly timeconsuming. Chen and Zhang [12] proposed FCM_S1
and FCM_S2, two variants of FCM_S algorithm in
order to reduce the computational time. These two
algorithms introduced the extra mean and medianfiltered image.
International Conference on Recent Advances in Mechanical Engineering and Interdisciplinary Developments [ICRAMID - 2014]
ISBN 978-1-4799-3158-3
m
ij
d2 (xi,vj)
(1)
(2)
(b+1)
International Conference on Recent Advances in Mechanical Engineering and Interdisciplinary Developments [ICRAMID - 2014]
vj =
ji(xi+
r)
m
ji
/ (1+a)
m
ji
||xi-vj||2 +
m
jr
||
-vj||2 (7)
vj =
ji(xi+a ) / (1+a)
m
ji
(9)
ISBN 978-1-4799-3158-3
(6)
(1-ukj)m||xj-vk||2
(10)
International Conference on Recent Advances in Mechanical Engineering and Interdisciplinary Developments [ICRAMID - 2014]
Jm =
ki
||xi-vk||2 + Gki]
uki = 1/
(12)
vk =
(13)
(a)
(b)
(c)
Fig.1. Comparison of segmentation results on synthetic test image.
(a) Original image, (b) FCM Result, (c) FLICM Result
ISBN 978-1-4799-3158-3
(b)
(11)
(a)
(c)
Fig.2. Comparison of segmentation results on test image.
(a) Original image with salt and pepper noise, (b) FCM Result,
(c) FLICM Result
V. CONCLUSION
In this paper, a novel robust fuzzy local
information c-means (FLICM) algorithm for image
clustering was introduced. The proposed algorithm
can detect the clusters of an image overcoming the
disadvantages of the known FCM algorithms and
their variants. This is achieved by introduces a new
factor as a local (spatial and gray) similarity measure
which aims to guarantee robustness both to noise and
outliers. This is also enforced by the way that spatial
and gray level image information are combined in the
algorithm; the factor combines in a fuzzy manner the
spatial and gray level information, rendering the
algorithm more robust to all kind of noises, as well as
to outliers.
REFERENCES
[1] X. Munoz, J. Freixenet, X. Cufi, and J. Marti, Strategies
for image segmentation combining region and boundary
information, Pattern Recognit. Lett., vol. 24, no. 1, pp.
375392, 2003.
[2] D. Pham, C. Xu, and J. Prince, A survey of current
methods in medical image segmentation, Annu. Rev
Biomed. Eng., vol. 2, pp. 315337, 2000.
[3] L. Zadeh, Fuzzy sets, Inf. Control, vol. 8, pp. 338
353, 1965.
[4] J. Udupa and S. Samarasekera, Fuzzy connectedness
and object definition: Theory, algorithm and applications in
image segmentation,Graph. Models Image Process., vol.
58, no. 3, pp. 246261, 1996.
[5] Y. Tolias and S. Panas, Image segmentation by a fuzzy
clustering algorithm using adaptive spatially constrained
membership functions,IEEE Trans. Syst., Man, Cybern.,
vol. 28, no. 3, pp. 359369, Mar.1998.
[6] J. Noordam, W. van den Broek, and L. Buydens,
Geometrically guided fuzzy C-means clustering for
multivariate image segmentation, in Proc. Int. Conf.
Pattern Recognition, 2000, vol. 1, pp. 462465.
International Conference on Recent Advances in Mechanical Engineering and Interdisciplinary Developments [ICRAMID - 2014]
ISBN 978-1-4799-3158-3