Image Segmentation using Fuzzy Clustering: A Survey

Samina Naz, Hammad Majeed

Department of Computer Science National University of Computer & Emerging Sciences Islamabad, Pakistan samina_853@yahoo.com, hammad.majeed@nu.edu.pk
Abstract This paper presents a survey of latest image segmentation techniques using fuzzy clustering. Fuzzy C-Means (FCM) Clustering is the most wide spread clustering approach for image segmentation because of its robust characteristics for data classification. In this paper, four image segmentation algorithms using clustering, taken from the literature are reviewed. To address the drawbacks of conventional FCM, all these approaches have modified the objective function of conventional FCM and have incorporated spatial information in the objective function of the standard FCM. The techniques that have been reviewed in this survey are Segmentation for noisy medical images with spatial probability, Novel Fuzzy C-Means Clustering (NFCM), Fuzzy Local Information C-Means (FLICM) Clustering Algorithm and Improved Spatial Fuzzy C-Means Clustering (ISFCM) algorithm. Keywords-image segemntation;clustering; fuzzy c-means; memebrship function; spatial probability;histogram;sparse 3D transform-domain collabrative filetring.

Humayun Irshad
Department of Computer Science Comsats Institute of Information Technology Islamabad, Pakistan humayun.irshad@gmail.com input data into groups or clusters on the basis of some similarity criterion, such that similar data objects belongs to the same group. Some examples of measures that can be used in clustering include distance, connectivity, and intensity. The resulting partition improves the understanding of human beings and helps in a more informed decision making. The advantage of using similarity measure is to control the clusters formation. To investigate cluster formation two types of clustering approaches are commonly used: One is Hard Clustering and the other is Soft/Fuzzy Clustering [44]. In Hard Clustering, data is divided into a number of distinct clusters, where each data element belongs to exactly one cluster, whereas fuzzy clustering is a process of assigning them membership levels, and then using these levels to assign data elements to one or more clusters or classes in the image/data set. In Soft/Fuzzy Clustering data elements can belong to more than one cluster with a degree of some membership values. Areas of applications of fuzzy cluster analysis include data analysis, pattern recognition and image segmentation. Figure 1 shows how fuzzy clustering is used in segmentation of images. In real applications where there is usually no sharp or crisp boundary between clusters, fuzzy clustering is often suited for classification of data in decision oriented applications like tissue classification, tumor detection etc. One of the most difficult tasks in image analysis and computer vision applications is to classify correctly the pixel values as there are no crisp boundaries between objects in an image. In order to address this diffucult task, fuzzy clustering techniques are proving to be a fruitful research area. The method of image segmentation using fuzzy clustering technique provides a mean of classifying pixel values with a great extent of acuracy. The advantage of Fuzzy systems is that they are easy to understand, since the membership functions partition the data space properly. Many approaches in the literature however do not have this property [38, 39, 40, 41, 45]. The present survey is an effort to explore image segmentation techniques that perform classification task using fuzzy clustering. Section III presents principal methodologies on image segmentation by fuzzy clustering as most researchers are interested in a more swift efficient and robust segmentation solution for a given problem. In Section IV we discuss quantitative performance analysis of these appraoches. In section V, we conclude this survey and give some future directions.

I.

INTRODUCTION

Image segmentation is one of the most widespread means to classify correctly the pixels of an image in a decision oriented applications. Image segmentation is a technique that partitions an image into uniform and non-overlapping regions based on some likeness measure [6]. This technique has a variety of applications including computer vision, image analysis, medical image processing, remote sensing and geographical information system. Image segmentation is based on two basic properties of image 1) intensity values involving discontinuity that refers to sudden or abrupt changes in intensity as edges and 2) similarity that refers to partitioning a digital image into regions according to some pre-defined likeness criterion. Much interest has been generated in recent years in the field of image analysis and is becoming more and more imperative. In this context, variety of segmentation techniques have been probed in the literature involving Histogram Thresholding [36], Edge Detection [37], Region Growing [46], Watershed Transformation [45], Clustering and Soft Computing techniques. Soft Computing techniques involve image segmentation using Genetic algorithms [38, 39, 40] Fuzzy Logic techniques [38, 41] and Neural Network based approaches [42, 43]. The cluster analysis is restricted to a statistical classification technique for discovering whether the individuals of a population falling into different groups by making quantitative evaluation of different distinctiveness. It divides

Input Image

Feature Space Transformation

segmentation since it has robust characteristics for ambiguity and can preserve much more information than hard clustering approaches [14]. FCM was proposed by Dunn [20] and later on it was modified by Bezdek [21]. FCM assigns pixels to each class by means of fuzzy membership function. Lets suppose X =(x1, x2, x3xN) denotes an image with N pixels to be categorized into C clusters. FCM is the iterative minimization of the following objection function: J=

Clustering in Feature Space

j =1 i =1

uijm || xj-vi||2

(1)

Cluster C1

Cluster Cn

where uij is the membership of pixel xj in ith cluster, vi is the ith cluster center, m is the fuzzifier that controls the fuzziness of resulting partitions and lies between 1 < m and ||.|| is a norm metric. Usually Euclidean distance between pixel xj and the center of ith cluster vi, is used as norm metric [22] and [23].The membership function and cluster centers are updated as:

(2)
Segment S11

Segment S1n

Segment S1n

Segment Snn
N

Figure1. Segmentation Using Clustering Adapted from [31]

II.

BACKGROUND

vi =

u
j =1 N j =1

ij

xj
(3)

Numerous methodologies have been proposed and a dense literature is available for extracting information from an image and to partition it into different regions. But all suffer from different limitations in terms of time complexity, accuracy. This is due to not well defined boundaries of clusters within the image, so techniques other than fuzzy result in disambiguates in segmented images, on the other hand fuzzy image segmentation methodologies yield good results. There is some sort of ambiguity associated with real-world images and thus segmenting such images results in fuzzy regions. Clustering methods use information like brightness and spatial location of pixels. These methods lack the ability to separate image regions having similar pixels intensities by considering only their pixel intensity. The pixels on an image are highly correlated, that is the pixels in the immediate neighborhood possess nearly the same feature data. Therefore, the spatial correlation of adjacent pixels is an important characteristic that can be of great aid in image segmentation. Image segmentation using fuzzy clustering has been a subject of much interest in recent years. The objective function based fuzzy clustering algorithms includes Fuzzy C-Means (FCM) algorithm, Gustafson-Kessel algorithm (GK), Gaussian Mixture Decomposition (GMD) algorithm, Fuzzy C-Varieties (FCV) algorithm, Adaptive Fuzzy-C varieties (AFC) algorithm, Fuzzy C-Shell (FCS) algorithm, Fuzzy C-Spherical Shells (FCSS) algorithm, Fuzzy C-Rings algorithm, Fuzzy C-Quadric Shells (FCQS) algorithm, Fuzzy C- Rectangular Shells (FCRS) algorithm and etc [1]. Among these above mentioned fuzzy clustering methods FCM is the most accepted means of image

ij

The cluster centers can either be initialized randomly or by an approximation method. On noisy images, FCM does not incorporate spatial information which makes it sensitive to noise and other image artifacts. Furthermore, as FCM cluster assignment is based exclusively on the distribution of pixel intensity, that makes it sensitive to intensity variations in the illumination or the geometry of the object [9]. To overcome the drawbacks of FCM a preprocessing image smoothing step has also been proposed in [24][25][26]. In literature several other algorithms have also been proposed in order to improve the robustness of conventional FCM. These can be broadly classified into two main categories: 1) imposing spatial constraints to clustering algorithms [7, 9, 10, 11] and 2) introducing dissimilarity index that is insensitive to intensity variations in the objective function of FCM [9, 11]. In the last few years spatial information is incorporated into original FCM algorithm in order to improve segmentation accuracy and has gained enormous significance. Sudhavani and Sathyaparsad [47] presented modified fuzzy C-Means clustering algorithm for lip image segmentation. Kannan et al. [12] have developed an efficient fuzzy segmentation algorithm for breast magnetic resonance imaging data. They have acquired an objective function of

FCM called Kernel Induced FCM which is based on hyper tangent function which is in turn based on two functions namely kernel, hyper tangent and Langrangian multipliers. Huynh Van Lung and Jong-Myon Kim [13] have developed generalized Spatial Fuzzy C-Means Clustering algorithm (GSFCM] for brain MRI segmentation. GSFCM utilizes given pixel attributes and spatial local information weighted equally to neighbors based on their distance attributes. Results have shown that GSFCM outperforms conventional FCM. Huiyu Zhou et al. [29] have presented a mean shift based FCM for the extraction of skin lesions. They have proposed a mean shift based fuzzy c-means objective function that is a mean field term is incorporated in the standard FCM objective function. Experimental results have shown that their algorithm is capable of extracting skin lesion borders proficiently. Ruoyu Du and Hyo Jong Lee [28] have proposed an enhanced segmentation technique that applies sigma filter to vary neighboring pixels of targets. Visual and quantitative assessment has revealed that the proposed method works better than the original FCM. III. FUZZY CLUSTERING SEGMENTATION
METHODOLOGIES FOR IMAGE

B. Fuzzy Logic Information C-Means (FLICM) Clustering Algorithm Krindis and Chatzis [3] have developed Fuzzy Logic Information C-Means Clustering (FLICM) algorithm to overcome the drawbacks in algorithms presently available in literature [16], [17] and [18]. FLICM is introduced with a new factor in objective function of FCM and the new factor Gki has following characteristics [3]: It incorporates local spatial and gray level information in a fuzzy way to preserve robustness and noise insensitiveness. It controls the influence of the adjacent pixels depending on their distance from middle pixel. It uses original image as input and avoids preprocessing steps to preserve image details. It is free from any parameter selection. This new novel fuzzy factor is defined mathematically as [3]:

Gki =
i j

1 (1 kj ) m || xj vk ||2 dij + 1 jNi

(4)

In this section we present a review of various image segmentation techniques using fuzzy clustering recently proposed in the literature. A. Robust segmentation for Noisy Medical Images Using Fuzzy Clustering with Spatial Probability S. Zulaikha Beevi and M. Mohammed Sathik [2] have developed a robust segmentation technique that exploit histogram based FCM algorithm for the segmentation of medical images. The algorithm first removes noise from the images and then segmentation is performed. Sparse 3D transform-domain collaborative filtering [27] is used to perform denoising. Histogram is used to initialize the parameters of the FCM to avoid convergence in local minima. Spatial probability is incorporated in the objective function to boost the robustness of algorithm against noise. In the proposed segmentation methodology, two types of spatial information are incorporated in the membership function of FCM. First is Apriori probability and other is Fuzzy Spatial Information. Apriori Probability is incorporated in the membership function to assign a noisy pixel to a cluster that contains a large number of noisy pixels in its neighborhood. Fuzzy Spatial Probability is incorporated in the membership function and a pixel gets a higher membership value to a cluster when its neighborhood pixels have high membership value to that cluster [2]. This approach converges more quickly than the conventional FCM and attains reliable segmentation accuracy apart from noise levels.

where ith pixel is the center of the local window e.g, 3*3, k is reference cluster and jth pixel belongs to the set of neighbors that fall into a window around ith pixel (Ni) and dij is Euclidean distance between pixels i and j [3]. The factor Gki is free from any control parameter for balancing image detail and noise. The stability between noise and image details is automatically attained by the fuzziness of every pixel of the image. The factor Gki formulates the influence of the pixels within the local window, to change flexibly according to their distance from the middle pixel by using Euclidean distance that is dij. It is important to note that the factor Gki reflects the damping extent of the neighbors with the spatial distances from the central pixel [3]. As compared to other parameters (a or ) presented in [24], [26] and [15] the factor Gki can vary adaptively with diverse spatial locations or distances from the middle pixel [3]. This factor is used in the objective function of the enhanced FCM algorithm as:

jm = [uki m || xi vk ||2 +Gki ]

i =1 k =1

(5)

The two essential conditions for Jm to be at its local minimum extreme with respect to uki and vk are defined as: (6)

vk =

u
i =1 N i =1

m ki

xi
(7)

m ki

After convergence defuzzification of partition matrix U is performed using maximum membership method as

Ci = arg K {max{uki }}; k = 1, 2,........, c

FLICM is proved to be more robust due to the introduction of Gki factor in objective function of FCM and it has characteristics like [3] noise insensitive preserves image details free from any parameter selection mechanism like a or requires no preprocessing steps on input image and thus can be applied directly on the input image In the absence of any prior noise knowledge, FLICM is an optimal strategy for clustering of images. C. Novel Fuzzy C-Means Clustering Algorithm (NFCM) Kannan et al. [4] have proposed a Novel Fuzzy Clustering C-Means Algorithm (NFCM) for intensity inhomogeneities or weighted bias estimation and segmentation of T1-T2 Brain MRI images of same patient. Authors also have presented a center knowledge method [4] to reduce the running time of the algorithm. The Center Knowledge Algorithm rearranges the data matrix with respect to its relabeling mean value and data is partitioned into c groups. A distance table is than created to show the distance between the elements within each group. Maximum distance between the groups is computed and a mean value is calculated. The objective function J has been modified for the said problem in [4] as follows:
C n C n jm = ik m [ dik 2 + (dik * )2 ] + [1 ik m ] c i =1 k =1 i =1

D. Improved Spatial Fuzzy C-Means Clustering Algorithm (ISFCM) Zulaikha et al. [5] have proposed an improved Spatial Fuzzy C-Means algorithm. The histogram based FCM (HFCM) is used to initialize the input parameters for ISFCM because HFCM converges more rapidly as it clusters the histogram of the image instead of clustering the whole image [5]. The spatial neighborhood information is incorporated into the standard FCM by a priori probability. The a priori probability is incorporated to show the spatial influence of neighboring pixels on the central pixel, which is automatically decided in the fuzzy membership algorithm [5]. Apriori probability assigns a noise pixel to that cluster to whose members are majority in the pixels neighborhood and fuzzy spatial information is used so that pixel gets higher membership values when their neighboring pixels have high membership values with the corresponding cluster [5]. The advantage of ISFCM is that it can overcome the noise sensitiveness of standard FCM. The incorporation of spatial information in the clustering process made the algorithm robust to noise and blurred edges. IV.
DISCUSSION

(8)

The methodologies discussed in section III are novel and robust since researchers have modified the objective function of conventional FCM algorithm and have incorporated spatial information in the objective function to cope with noise and other image artifacts. The segmentation algorithm for noisy medical images using fuzzy clustering with spatial probability proposed in [2] uses sparse 3D Transform-Domain collaborative Filtering for denoising input image [27]. The denoising algorithm linearly depends on the size of the input image and takes O (|X|) time [27]. Authors have evaluated quality of results by Segmentation Accuracy. Algorithm is tested on different noise levels of Additive White Guassian Noise (AWGN) and experimental results presented in [2] shows that their algorithm has outperformed the conventional FCM. Maximum segmentation accuracy of technique presented in [2] is 95%. Experimental results presented in [3] show that the FLICM algorithm is more robust when applied to synthetic images corrupted by Guassian, uniform and salt & pepper noise of different levels and the segmentation accuracy of the algorithm presented in [3] is very high as compared to techniques like FCM_S1 [25], FCM_S2 [25], EnFCM [26], FGFCM_S1 [15], FGFCM_S2 [15], FGFCM [15], K-means [32], SLINK [33]. Comparison Score [15] and [19] also depicts the robustness of FLICM. Also the new factor Gki guarantees the robustness of the algorithm to noise and outliers [3]. An important characteristic of FLICM is that it doesnt involve any preprocessing steps or algorithm for denoising of the image and is free of any parameter selection mechanism. Maximum segmentation accuracy of FLICM claimed in [3] is 99.98 %. Novel FCM (NFCM) algorithm revealed in [4] deals with intensity inhomogeneties of MRI brain images and it has been claimed in [4] that the NFCM approach has outperformed the

Where

dik 2 =|| xk vi ||2 And dik * =|| xk + bk vi ||2

(9)

(0,1)

And

, > 0

Objective function is minimized by employing same approach as that of conventional FCM. Jm is differentiated with respect to ik ,vi and bk and is than set equal to zero in order to get estimators for U, V and b. These estimators are then used to design an algorithm to calculate tissue class and bias class. Lagrange multiplier is then used to perform the task of membership evaluation as follows [4]:

Lm = ik m [ dik 2 + (dik * ) 2 ] +
i =1 k =1 C C n [1 ik m ] + [1 ik ] c i =1 i =1

(10)

Lm is then differentiated with respect to compute

ik and is set to zero to

ik *.

The advantage of NFCM is that it can be applied at an early phase of automated data analysis. NFCM is found to deal effectively with image intensity in-homogeneities and noise present in the image.

IFS [34] and KFCM [35]. Segmentation accuracy is computed using Silhouette value [30]. Silhouette values measures the degree of confidence in assigning an observation to a cluster. Maximum accuracy of NFCM found in [4] is 73%. ISFCM [5] and approach of [2] is more or less same as both techniques have employed the Histogram FCM (HFCM) to generate initial parameters for FCM and spatial information is incorporated in the objective function. In ISFCM [5] denoising is not performed separately and accuracy of their methodology is near 95%. The key advantage of ISFCM is its simplicity and thus it can be applied to large datasets. V.
CONCLUSION AND FUTURE DIRECTIONS

[9]

[10]

[11]

[12]

[13]

Segmentation is an important step in advance image analysis and computer vision and therefore is an ongoing research area although a dense literature is available. The incorporation of spatial information in to the objective function of standard FCM yields successful results for robust and effective image segmentation of noisy images [2, 3, 4, 5] and the techniques like [2], ISFCM [5] and NFCM [4] can be applied to segment colored images. The techniques [1, 2, 3 and 4] reviewed in this survey are applicable to analysis of MRI images and in future can be applied to other medical image types like CT and PET for better analysis. 3D volume of MR data based on segmentation using fuzzy clustering can be reconstructed and lesion volume can also be analyzed quantitatively. Furthermore in future a hybrid technique based on clustering algorithms and classifiers like Neural Networks and etc can be combined to work on input data set for better results and previously designed algorithm can be modified to work for color image segmentation. REFERENCES
[1] [2] www.wikipedia.com S Z Beevi and M.M. Sathik, A robust Segmentation Approach for Noisy Medical Iamges Using Fuzzy Clustering With Spatial Probability, European Journal of Scientific Research, Vol. 41 No. 3 pp. 437-451, 2010 S.Kirindis and V. Chatzis, A Robust Fuzzy Local Information CMeans Clustering Algorithm, IEEE Trans. Image Process. Vol.19, No. 5 , pp. 1328-1337,May 2010 S.R.Kannan, S. Ramathilagam, R.Pandiyarajan and A.Sathya, Fuzzy Clustering Approach in Segmentation of T1-T2 brain MRI, Int. Journal of Recent Trends in Engineering, Vol2, No.1, pp. 157-160, Nov.2009 S.Z.Beevi, M.M. Sathik and K.senthamarakannan, A Robust fuzzy Clustering Technique with Spatial neighborhood Information for Effective Medical Image Segmentation, Int. Journal of Computer Science and Information Security, Vol 7. No. 3, pp. 132-138, March 2010 G.Dong., M.Xie, Color clustering and learning for image segmentation based on neural networks, IEEE Transactions on Neural Networks, Vol.16, No.4, pp. 925936, 2005. Y.A. Tolias and S.M. Panas, "Image segmentation by a fuzzy clustering algorithm using adaptive spatially constrained membership functions", IEEE Transactions on Systems, Man, Cybernetics, Part A, Vol. 28, No. 3, pp. 359-369, 1998. J.Yu and Y. Wang, "Molecular Image Segmentation Based on Improved Fuzzy Clustering", Journal of Biomedical Imaging, Vol.2007, No.1, 9 pages, 2008.

[14]

[15]

[16]

[17] [18] [19]

[20]

[21] [22]

[23]

[3]

[24]

[4]

[25]

[5]

[26]

[6]

[27]

[7]

[28]

[29]

[8]

Liew AWC, Yan H, Law N-F., Image segmentation based on adaptive cluster prototype estimation, IEEE Transactions on Fuzzy Systems, Vol. 13, No.4, pp. 444453, 2005. Ahmed M.N., Yamany S.M., Mohamed N., Farag A.A., Moriarty T.,"A modified fuzzy c-means algorithm for bias field estimation and segmentation of MRI data, "IEEE Transactions on Medical Imaging, Vol. 21, No. 3, pp.193199,2002. Chen S. and Zhang D.," Robust image segmentation using FCM with spatial constraints based on new kernel-induced distance measure, "IEEE Transactions on Systems, Man, and Cybernetics, Vol. 34, No. 4, pp.19071916, 2004. Kannan, Ramathilagam and Sathya, "Robust Fuzzy C-Means in Classifying Breast Tissue Regions", In proceedings of ARTCOM International Conference on Advances in Recent Technologies in Communication and Computing, pp.543-545,2009 Huynh Van Lung and Jong-Myon Kim, "A generalized spatial fuzzy Cmeans algorithm for medical image segmentation", In proceedings of the 18th International conference on Fuzzy Systems, pp.409-414, 2009. D. Pham, An adaptive fuzzy C-means algorithm for image segmentation in the presence of intensity inhomogeneities, Pattern Recognit.Lett., vol. 20, pp. 5768, 1999. W. Cai, S. Chen, and D. Zhang, Fast and robust fuzzy c-means clustering algorithms incorporating local information for image segmentation, Pattern Recognit., vol. 40, no. 3, pp. 825838, Mar. 2007. R. Hathaway, J. Bezdek, and Y. Hu, Generalized fuzzy c-means clustering strategies using L norm distance, IEEE Trans. Fuzzy Syst., vol.8, pp. 576582, Oct. 2000 K.Wu and M. Yang, Alternative c-means clustering algorithms, Pattern Recognit., vol. 35, no. 10, pp. 22672278, 2002 J. Leski, Toward a robust fuzzy clustering, Fuzzy Sets Syst., vol. 137,no. 2, pp. 215233, 2003. F. Masulli and A. Schenone, A fuzzy clustering based segmentation system as support to diagnosis in medical imaging, Artif. Intell. Med.,vol. 16, no. 2, pp. 129147, 1999. J.C. Dunn, "A fuzzy relative of the ISODATA process and its use in detecting compact well separated clusters", Journal of Cybernetics, Vol.3, No.3, pp. 32-57, 1974. J.C. Bezdek, Pattern Recognition with Fuzzy Objective Function Algorithms, New York, Plenum Press, 1981. M. Ameer Ali, Gour C Karmakar and Laurence S Dooley, Image Segemntation Using Fuzzy Clustering Incorporating Spatial Information Keh-Shih Chuang, Hong-Long Tzeng, Sharon Chen, Jay Wu, Tzong-Jer Chen., Fuzzy c-means clustering with spatial information for image segmentation Computerized Medical Imaging and Graphics, Eleseiver, 2005 M. Ahmed, S. Yamany, N. Mohamed, A. Farag, and T. Moriarty, A modied fuzzy C-means algorithm for bias eld estimation and segmentation ofMRI data, IEEE Trans.Med. Imag., vol. 21, pp. 193199, 2002. S. Chen and D. Zhang, Robust image segmentation using FCM with spatial constraints based on new kernel-induced distance measure, IEEE Trans. Syst., Man, Cybern., vol. 34, pp. 19071916, 2004. L. Szilagyi, Z. Benyo, S. Szilagyii, and H. Adam, MR brain imagesegmentation using an enhanced fuzzy C-means algorithm, in Proc. 25th Annu. Int. Conf. IEEE EMBS, 2003, pp. 1721 K. Dabov, A. Foi, V. Katkovnik, and K. Egiazarian, "Image denoising by sparse 3D transform-domain collaborative filtering," IEEE Transactions on Image Processing, vol. 16, no. 8, August 2007. Ruoyu Du and Hyo Jong Lee, "A modified-FCM segmentation algorithm for brain MR images", In proceedings of ACM International Conference on Hybrid Information Technology, pp.25-27, 2009. Huiyu Zhou, Gerald Schaefer, Abdul Sadka and Emre Celebi, "Anisotropic mean shift based fuzzy c-means segmentation of skin lesions", Proceedings of the 5th international conference on Soft computing as transdisciplinary science and technology, pp.438-443, 2008.

[30] Kannan S R, A new segmentation system for brain MR images based on fuzzy techniques, Applied Soft Computing, 8 (4), 1599-1606, 2008. [31] Image Segmentation Using Clustering by Juraj Horvth [32] J. MacQueen, Some methods for classication and analysis of multivariate observations, in Proc. 5th Berkeley Symp. [33] R. Sibson, SLINK:An optimally efcient algorithmfor the single-link cluster method, The Comput. J., vol. 16, no. 1, pp. 3034, 1973. [34] Kapur, T., Grimson, W .E.L., Wells, W.M., Kikinis, R., Segmentation of brain tissue from magnetic resonance images, Medical Image Analysis, 1, 109127, 1996. [35] Brinkmann, B.H., Manduca, A., Robb, R.A, Optimized homomorphic unsharp masking for MR grayscale inhomogeneity correction, IEEE Transactions Medical Imaging 17, 161171, 1998. [36] S. K. Pal, R. A. King, and A. A. Hashim, Automatic grey level thresholding through index of fuzziness and entropy, Pattern Recognit.Lett., vol. 1, pp. 141146, Mar. 1983. [37] Batista, J., Freitas, R, An adaptive gradient-based boundary detector for MRI images of the brain, Image Processing And Its Applications, Seventh International Conference, vol.1, pp.440 444, Jul 1999. [38] M. Abdulghafour, Image segmentation using Fuzzy logic and genetic algorithms, Journal of WSCG, vol.11, no. 1, 2003 [39] Xian Bin Wen, Hua Zhang and Ze Tao Jiang, Multiscale Unsupervised Segmentation of SAR Imagery Using the Genetic Algorithm, Sensors, vol.8, 2008, pp.1704-1711. [40] George Karkavitsas and Maria Rangoussi,Object Localization in Medi cal Images Using Genetic Algorithms, International Journal of Signal Processing, 2005, pp.204-207. [41] Kanchan Deshmukh and G. N. Shinde, An adaptive neuro-fuzzy system for color image segmentation, J. Indian Inst. Sci., vol. 86, Sept.-Oct. 2006, pp.493-506. [42] Jander Moreira and Luciano Da Fontoura Costa, Neural-based color image segmentation and classication using self-organizing maps, Anais do IX SIBGRAPI, 1996, pp.47-54. [43] Metin Kaya, Image Clustering and Compression Using An Annealed Fuzzy Hopeld Neural Network, International Journal of Signal Processing, 2005, pp.80-88. [44] www.reference.wolfram.com [45] A. Bleau and L. J. Leon, Watershed-based segmentation and region merging, CVIU, vol. 77, no. 3, pp. 317370, Mar. 2000. [46] J.P. Fan, D.K.Y Yau, and A.K. Elmagarmid. Automatic image segmentation by integrating color edge extraction and seeded region growing. IEEE Transactions on Image Processing, 10(10):14541466, 2001 [47] Sudhavani and Sathyaprasad, "Segmentation of Lip Images by Modified Fuzzy C-means Clustering Algorithm", IJCSNS International Journal of Computer Science and Network Security, Vol.9 No.4, pp.187-191, April 2009.