This action might not be possible to undo. Are you sure you want to continue?

Images

Manimala Singha

Dept. of Computer Science

Assam University Silchar

India. Pin code 788011

n.manimala888@gmail.com

K. Hemachandran

Dept. of Computer Science

Assam University Silchar

India. Pin code 788011

khchandran@rediffmail.com

Abstract— This paper introduces efficient and fast algorithms for unsupervised image segmentation,

using low-level features such as color, applied on satellite images. With the increase in spatial resolution

of satellite imagery, the image segmentation technique for generating and updating geographical

information are becoming more and more important. The present paper describes a satellite image

segmentation technique using M-band fuzzy c-Means features. In remotely-sensed multispectral imagery

the variations in the reflectivity of surface materials across different spectral bands provide a

fundamental mechanism for understanding the image features. Fuzzy methods in remote sensing have

received growing interest for their importance in situations where the geographical phenomena are

inherently fuzzy. The proposed approach is based on that first enhance multispectral image and then

applying clustering technique, using La*b* color space and the vectors are used as inputs for the k-means

or fuzzy c-means clustering methods, for a segmented image whose regions are distinct from each other

according to color and texture characteristics.

Keywords- Image segmentation; k-means; fuzzy c-means (FCM); La*b* color space.

Introduction

Satellite image segmentation has been a focused research area in the image processing, for the last few

decades. Many papers has been published, mainly focused on gray scale images and less attention on color

image segmentation, which convey much more information about the object or images. Image segmentation is

typically used to locate objects and boundaries in images. Image segmentation is a process of partitioning image

pixels based on selected image features. The pixels that belong to the same region must be spatially connected

and have the similar image features. If the selected segmentation feature is color, an image segmentation process

would separate pixels that have distinct color feature into different regions, and, simultaneously, group pixels

that are spatially connected and have the similar color into the same region [10]. In color imagery, image pixels

can be represented in a number of different color spaces e.g. RGB, HSV or LUV [5, 9, and 10].

The CIELAB color space, adopted as an international standard in the 1970’s, provides perceptually

uniform space, which means the Euclidean distance between two color points in the CIELAB color space

corresponds to the perceptual difference between the two colors by the human vision system [1]. This

property has made the CIELAB color space to be attractive and useful for color analysis, and the CIELAB

color space has shown its superior performance in many color image applications ([2]-[4]). Based on these

reports, the CIELAB color apace has been chosen for color clustering.

High-resolution images are useful for applications such as transportation network mapping, disaster

preparedness, urban planning, precision farming, and telecommunications. On the other hand, low-resolution

satellite images, with ground resolution greater than 10m, are useful for applications like environmental

assessment, regional mapping, forestry management, widespread disaster assessment, and urban monitoring. It

is often necessary to enhance multispectral radiance or reflectance data to create an image that is suitable for

visual interpretation.

Manimala Singha et al. / International Journal on Computer Science and Engineering (IJCSE)

ISSN : 0975-3397 Vol. 3 No. 12 Dec 2011 3756

Clustering is an unsupervised, computationally efficient and simplest technique, which can be applied to

multi dimensional data and the results are meaningful only if the homogenous non-textured color regions define

the image data, in general ([5]-[17]). The pixel-based segmentation technique, consisting of Fuzzy C Means [6]

and K-Means [14] considers only the spectral pattern to segment the image. These techniques are not sufficient

to segment high-resolution satellite images due to the variability of spectral and structural information in such

images.

This paper considers the segmentation problem of image regions based on two clustering methods: color

feature similarity using fuzzy c-means or k-means. The rest of the paper is organized as follows: In section I, a

brief description the clustering techniques presented. The section II describes the Image segmentation using

kmeans (KM). The section III, Image segmentation using fuzzy C- means (FCM). The experimental results are

given in section IV. Finally the conclusions are presented in section V.

I. CLUSTERING TECHNIQUE ON IMAGES

Clustering is considered to be the most important unsupervised learning problem. It deals with finding a

structure in a collection of unlabeled data. Clustering is the process of organizing objects into groups whose

members are similar in some way. A cluster is therefore a collection of objects which are “similar” between

them and are “dissimilar” to the objects belonging to other clusters. The two basic clustering techniques are K-

means (an exclusive clustering algorithm) and Fuzzy C-means (an overlapping clustering algorithm). ([5], [20] -

[23]).

A. K-Means Algorithm

K-means algorithm was originally introduced by McQueen in 1967 [22]. It is a non-fuzzy clustering method

whereby each pattern can only belong to one centre at any one time. The K-means algorithm is an iterative

technique that is used to partition an image into K clusters.

Let X = {x

1

, x

2

... x

n

} represent a set of pixels of the given image, where n is the number of pixels. V = {v

1

,

v

2

... v

k

} is the corresponding set of cluster centres, where k is the number of clusters. The aim of K-means

algorithm is to minimize the objective function J(V), in this case a squared error function:

}(v) = ∑ ∑ [x

Ij

−v

j

[

2

k

t

j=1

k

I=1

(1)

Where, [x

Ij

−v

j

[ is the Euclidean distance between x

ij

and v

j

.k

i

is the number of pixels in the cluster i.

The difference is typically based on pixel colour, intensity, texture, and location, or a weighted combination of

these factors. In our study, we have considered pixel intensity.

The i

th

cluster centre v

i

can be calculated as:

:

ì

=

1

k

i

∑ x

ì]

k

t

]=1

(2)

for i = 1, ..., k.

The basic algorithm is:

i) Randomly select k cluster centres.

ii) Calculate the distance between all of the pixels in the image and each cluster centre.

iii) A pixel is assigned to a cluster based on the minimum distance.

iv) Recalculate the centre positions using equation (2).

v) Recalculate the distance between each pixel and each centre.

vi) If no pixel was reassigned, then stop, otherwise repeat step (iii).

This algorithm is guaranteed to converge, but it may not return the optimal solution. The quality of the

solution depends on the initial set of clusters and the value of k.

B. Fuzzy C-means (FCM) Algorithm

Fuzzy clustering is one of the most widely used fuzzy approaches in image segmentation. The FCM

algorithm was originally introduced by Bezdek in 1981 [23]. It is an iterative algorithm. FCM can be used to

build clusters (segments) where the class membership of pixels can be interpreted as the degree of

belongingness of the pixel to the clusters.

Manimala Singha et al. / International Journal on Computer Science and Engineering (IJCSE)

ISSN : 0975-3397 Vol. 3 No. 12 Dec 2011 3757

Let X = {x

1

, x

2

... x

n

} represent a set of pixels of the given image, where n is the number of pixels and V =

{v

1

, v

2

... v

c

} is the corresponding set of fuzzy cluster centers, where c is the number of clusters. The main aim is

to minimize the objective function J(U,V), which is a squared error clustering criterion defined as :

}(0, v) = ∑ ∑ μ

Ij

m c

j=1

[x

I

−v

j

[

2

n

I=1

(3)

Where, [x

Ij

−v

j

[ is the Euclidean distance between x

ij

and v

j

. µ

ij

is the membership degree of pixel x

i

to the

cluster centre v

j

and μ

ij

has to satisfy the following conditions:

μ

Ij

∈ |u,1], , ∀i = 1,...n, ∀j = 1,...c (4)

∑ μ

Ij

c

j=1

= 1 , ∀i = 1,...n (5)

0 = (μ)

Ij

n∗c

is a fuzzy partition matrix. Parameter m is called the “fuzziness index”; it is used to control the

fuzziness of membership of each pixel. The value of m should be within the range m ∈ [1, ∞]. m is a weighting

exponent that satisfies m >1 and controls the degree of “fuzziness” in the resulting membership functions: As

m approaches unity, the membership functions become more crisp, and approach binary functions. As m

increases, the membership functions become increasingly fuzzy.

The FCM algorithm can be performed by the following steps:

i) Initialize the cluster centres V = {v

1

, v

2….,

v

c

}, or initialize the membership matrix μ

ij

with random

value such that it satisfies conditions (4) and (5). Then calculate the cluster centres.

ii) Calculate the fuzzy membership μ

ij

using:

p

ì]

=

1

∑ (

d

i]

d

ik

)

2

m-1

c

k=1

(6)

Where, d

ij

= ||x

i

− v

j

||, ∀i = 1... n, ∀j =1,...,c.

iii) Compute the fuzzy centres v

j

using:

:

]

=

∑ (µ

i]

)

m

x

i

n

i=1

∑ (µ

i]

)

m n

i=1

(7)

iv) Repeat step (ii) to (iii) until the minimum J value is achieved.

II. PROPOSED IMAGE SEGMENTATION TECHNIQUE

A. K-MEANS(KM)

• Read color image.

• Enhance the image

• Convert the image into La*b* color space.

• Segment the feature vector using K-Means Clustering.

• Label every pixel in the image using the Results from K-Means Clustering.

• Create segment image.

B. Fuzzy C –means(FCM)

• Read color image.

• Enhance the image

• Convert image into La*b* color space.

• Segment the feature vector using FCM.

• Assign the pixels to the clusters as follows:

Manimala Singha et al. / International Journal on Computer Science and Engineering (IJCSE)

ISSN : 0975-3397 Vol. 3 No. 12 Dec 2011 3758

o A pixel x is assigned to cluster i such that the value of the membership function of x for i is

maximum.

• Label every pixel in the image using the results from FCM.

• Create segment image.

III. EXPERIMENTAL RESULTS

The algorithms are implemented on Matlab 7.3 (The Mathworks Inc.). The first test satellite images the LAN

file, paris.lan, contains a 7-band 512x512 Landsat image (called test1). A 128-byte header is followed by the

pixel values, which are band interleaved by line (BIL) in order of increasing band number. They are stored as

unsigned 8-bit integers, in little-endian byte order [MATLAB 7.3]. The near infrared (NIR) band is important

because of the high reflectance of chlorophyll in this part of the spectrum. It is even more useful when

combined with visible red and green. To read multispectral bands from the LAN file using the MATLAB

function multibandread. Here we are reading band through multibandread are infrared, red, green. To enchance

multispectral image is to use a decorrelation stretch, which enhances color separation across highly correlated

channels. Use MATLAB Functrion decorrstretch to perform the decorrelation. The surface features have

become much more clearly visible. The spectral differences across the scene have been exaggerated. The green

area is the Bois de Boulogne, a large park on the western edge of Paris. A property of color infrared composites

is that they look red in areas with a high vegetation (chlorophyll) density, right red on the left edge, a large park

(the Bois de Boulogne) located west of central Paris within a bend of the Seine River. The second satalite image

downloaded from [25] of size 1024x1024 (called test2). To segment the multispectral images, the proposed

algorithms K-Means and Fuzzy C-Means, have been implemented and tested on satellite images. The KM and

FCM algorithms based on color segmentation, as described in section IV, have been tested on La*b* color space

and the results are shown in fig. 2(a) and 2(b). The number of clusters were chosen, in such a manner, to

segment the region of interest completely, for the image provided. The time taken to segment the satellite

test image by different algorithms is given in Table 1. From the results, it can be conclude that the segmentation

time (table 1.) for FCM method is more and the quality of the segmentation is much accurate than the KM

method (fig.2a and 2b.), for the same test satallite image. This is in conformity with the results of [24].

TABLE1. Segmentation time of different algorithms on a test images:

Algorithms Time in sec.

K-Means test1 4.13

Fuzzy C-Means test1 5.65

K-Means test2 4.01

Fuzzy C-Means test2 5.30

Manimala Singha et al. / International Journal on Computer Science and Engineering (IJCSE)

ISSN : 0975-3397 Vol. 3 No. 12 Dec 2011 3759

(a)

(b)

Manimala Singha et al. / International Journal on Computer Science and Engineering (IJCSE)

ISSN : 0975-3397 Vol. 3 No. 12 Dec 2011 3760

(c)

(d)

Fig2. Color image segmentation using (a), (c) K-means and (b), (d) FCM.

IV. CONCLUSIONS

Algorithms for segmentation of satellite images using Fuzzy C-Means (FCM), K-Means (KM) as described in

section II, have been implemented and tested on satellite image. Before doing image segmentation some low-

resolution satellite images, it is often necessary to enhance multispectral image that is suitable for visual

interpretation, for that we have used decorrelation stretch technique. So the surface features have become much

more clearly visible from the figures we can see that. The experimental result is shown in fig 2. From the

Manimala Singha et al. / International Journal on Computer Science and Engineering (IJCSE)

ISSN : 0975-3397 Vol. 3 No. 12 Dec 2011 3761

Table.1 it seen that KM segmentation methods are fast as compared to FCM, but segmentation methods using

FCM give better segmented images with finer details and accurate location.

REFERENCES

[1] G. Wyszecki and W.S. Stiles, “Color science : concepts and methods, quantitative data and formulae”. John Wiley and Sons, 2nd

edition, 2000.

[2] L. Jin and D. Li, “A Switching vector median based on the CIELAB color space for color image restoration”, Signal Processing, Vol.

87, pp.1345-1354, 2007.

[3] H. M. Lotfy and A. S. Elmaghraby, “CoIRS: Cluster-oriented Image Retrieval System”, Proc. 16th IEEE International Conference on

Tools with Artificial Intelligence (ICTAI), pp. 224-231, 2004.

[4] N. J. Kwak, D. J. Kwon, Y. G. Kim, and J. H. Ahn, “Color image segmentation using edge and adaptive threshold value based on the

image characteristics”, Proc. International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS), pp.

255-258. 2004.

[5] G. J. Klir and B. Yuan, Fuzzy Sets and Fuzzy Logic-Theory and Applications, PHI. 2000.

[6] J. C. Bezdek, R. Ehrlich, and W. Full, “FCM: The Fuzzy c-Means clustering algorithm”, Computers and Geosciences 10, 191–203,

1984.

[7] T. Martin, “A fuzzy clustering technique for medical image segmentation”, Proceedings of international symposium on

evolving fuzzy systems, September 2006, pp. 118-122.

[8] M. Li and R. C. Stuanton, “A modified fuzzy c-means image segmentation algorithm for use with uneven illumination patterns”,

Pattern recognition, vol. 40, 2007, pp. 3005-3011.

[9] K. Chuang, H. Tzeng, S. Chen, J. Wu, and T. Chen, “Fuzzy c-means clustering with spatial information for image segmentation”,

Computerized Medical Imaging and Graphics, vol. 30, 2006, pp.9–15.

[10] T. Q. Chen and Y. Lu, “Color image segmentation an innovative approach”, Pattern recognition, vol. 35, 2002, pp. 395-405.

[11] Y. Yang, C. Zheng, and P. Lin, “Fuzzy c-means clustering algorithm with a novel penalty term for image segmentation” Opto-

electronic review, Vol.13, Issue 4, 2005, pp. 309-315.

[12] Y. A. Tolias and M. P. Stavros, “Image segmentation by a fuzzy clustering algorithm using adaptive spatially constrained

functions”, IEEE Transactions on Systems, Man, and Cybernetics, Vol. 28, Issue 3. May 1998, pp. 359-369.

[13] J. B. McQueen,"Some methods of classification and analysis of multivariate observations", Proceedings of Fifth Berkeley Symposium

on Mathematical Statistics and Probability, Berkeley, 1967.

[14] S. Khan and A. Ahmad, “Cluster center initialization algorithm for K-Means clustering”, Pattern Recognition Letters, 2004, 25(11):

1293-1302.

[15] J. A. Hartigan and M. A. Wong, “A K-means clustering algorithm”, Appl. Stat. 28(1), 100–108, 1979.

[16] T. W. Chen, Y. L. Chen, and S. Y. Chien, “Fast image segmentation based on K-Means clustering with histograms in HSV color

space,” Proceedings of IEEE International Workshop on Multimedia Signal Processing (MMSP2008), Cairns, Queensland, Australia,

Oct 2008, pp. 322–325.

[17] M. Anirban and M. Ujjwal. “Unsupervised Pixel Classification in Satellite Imagery Using Multiobjective Fuzzy Clustering Combined

With SVM Classifier”, IEEE Transactions On Geoscience And Remote Sensing, Vol. 47, No. 4, April 2009.

[18] M. Antonini, M. Barlaud, P Mathieu, and I. daubechies,” Image Coding using Wavelet Transform”, IEEE transaction on image

processing, Vol.1, no. 2. April 1992.pp-205-220.

[19] M. L. Kherfi, D. Ziou, and A. Bernardi, Laboratories Universities Bell,” Image Retrieval from the World Wide Web: Issues,

Techniques, and Systems” ,In, ACM Computing Surveys, Vol. 36, No. 1, March 2004, pp. 35–67.

[20] R. C. Gonzalez and R. E. Woods. “Digital Image Processing”, 2nd ed., Pearson Education, 2000.

[21] R. C. Gonzalez R. E. Woods, and S. L. Eddins, “Digital Image Processing using MATLAB”, Pearson Education, 2008.

[22] J. B. MacQueen, "Some Methods for classification and Analysis of Multivariate Observations", Proceedings of 5th Berkeley

Symposium on Mathematical Statistics and Probability. University of California Press. 1967 pp. 281–297.

[23] J.C. Bezdek,‟Pattern Recognition and Fuzzy Objective Function Algorithms”, Plenum Press, New York, 1981.

[24] E. A. Zanaty, “Image Segmentation Technique in Medical Sciences”. ScienceEchoes 2006,7,14-22.

[25] http://sipi.usc.edu/database/database.php

AUTHORS PROFILE

Ms. Manimala Singha received her B.Sc. and M.Sc. degrees in Computer

Science from Assam University, Silchar in 2005 and 2007 respectively.

Presently she is working, for her Ph.D., as a Research Scholar and her area of

interest includes image segmentation, feature extraction, and image searching in

large databases.

Prof. K. Hemachandran is associated with the Dept. of Computer Science,

Assam University, Silchar, since 1998. He obtained his M.Sc. Degree from Sri

Venkateswara University, Tirupati and M.Tech. and Ph.D. Degrees from Indian

School of Mines, Dhanbad. His areas of research interest are Image Processing,

Software Engineering and Distributed Computing.

Manimala Singha et al. / International Journal on Computer Science and Engineering (IJCSE)

ISSN : 0975-3397 Vol. 3 No. 12 Dec 2011 3762

In our study. and location. Image segmentation using fuzzy C. / International Journal on Computer Science and Engineering (IJCSE) Clustering is an unsupervised. B. . V = {v1. ISSN : 0975-3397 Vol.Manimala Singha et al. The section II describes the Image segmentation using kmeans (KM). x2.. v2. texture. v) Recalculate the distance between each pixel and each centre. The ith cluster centre vi can be calculated as: = ∑ for i = 1. (2) ii) Calculate the distance between all of the pixels in the image and each cluster centre. computationally efficient and simplest technique. This paper considers the segmentation problem of image regions based on two clustering methods: color feature similarity using fuzzy c-means or k-means. consisting of Fuzzy C Means [6] and K-Means [14] considers only the spectral pattern to segment the image. K-Means Algorithm K-means algorithm was originally introduced by McQueen in 1967 [22]. Clustering is the process of organizing objects into groups whose members are similar in some way. iii) A pixel is assigned to a cluster based on the minimum distance. This algorithm is guaranteed to converge. [20] [23]).. A cluster is therefore a collection of objects which are “similar” between them and are “dissimilar” to the objects belonging to other clusters. These techniques are not sufficient to segment high-resolution satellite images due to the variability of spectral and structural information in such images. FCM can be used to build clusters (segments) where the class membership of pixels can be interpreted as the degree of belongingness of the pixel to the clusters. It is a non-fuzzy clustering method whereby each pattern can only belong to one centre at any one time. x − v is the Euclidean distance between xij and vj. iv) Recalculate the centre positions using equation (2). The basic algorithm is: i) Randomly select k cluster centres.. The pixel-based segmentation technique.. Let X = {x1. intensity.means (FCM). The aim of K-means algorithm is to minimize the objective function J(V). I. The two basic clustering techniques are Kmeans (an exclusive clustering algorithm) and Fuzzy C-means (an overlapping clustering algorithm).. vi) If no pixel was reassigned.. It deals with finding a structure in a collection of unlabeled data. but it may not return the optimal solution. The FCM algorithm was originally introduced by Bezdek in 1981 [23]. Finally the conclusions are presented in section V. 3 No. The difference is typically based on pixel colour.. a brief description the clustering techniques presented. or a weighted combination of these factors.ki is the number of pixels in the cluster i. ([5]. The experimental results are given in section IV. we have considered pixel intensity. The rest of the paper is organized as follows: In section I. The section III. CLUSTERING TECHNIQUE ON IMAGES Clustering is considered to be the most important unsupervised learning problem. otherwise repeat step (iii). in this case a squared error function: J(V) = ∑ ∑ x −v (1) Where. k. in general ([5]-[17]). Fuzzy C-means (FCM) Algorithm Fuzzy clustering is one of the most widely used fuzzy approaches in image segmentation. where n is the number of pixels. where k is the number of clusters. xn} represent a set of pixels of the given image. A. 12 Dec 2011 3757 . vk} is the corresponding set of cluster centres. The quality of the solution depends on the initial set of clusters and the value of k. The K-means algorithm is an iterative technique that is used to partition an image into K clusters. then stop. It is an iterative algorithm. which can be applied to multi dimensional data and the results are meaningful only if the homogenous non-textured color regions define the image data.

x2. The FCM algorithm can be performed by the following steps: i) Initialize the cluster centres V = {v1. As m increases. Create segment image. ∞]. ∀j =1. PROPOSED IMAGE SEGMENTATION TECHNIQUE B... n. Enhance the image Convert image into La*b* color space. m is a weighting exponent that satisfies m >1 and controls the degree of “fuzziness” in the resulting membership functions: As m approaches unity. or initialize the membership matrix μij with random value such that it satisfies conditions (4) and (5). xn} represent a set of pixels of the given image. 12 Dec 2011 3758 . which is a squared error clustering criterion defined as : J(U. Fuzzy C –means(FCM) Read color image. Label every pixel in the image using the Results from K-Means Clustering. 3 No.. x − v is the Euclidean distance between xij and vj.1 . Then calculate the cluster centres. vc}. ∀j = 1.V). it is used to control the U = (μ) ∗ fuzziness of membership of each pixel. Parameter m is called the “fuzziness index”. vc} is the corresponding set of fuzzy cluster centers. / International Journal on Computer Science and Engineering (IJCSE) Let X = {x1. ii) Calculate the fuzzy membership μij using: = ∑ ( ) (6) Where.n (5) is a fuzzy partition matrix... where n is the number of pixels and V = {v1. .. Segment the feature vector using K-Means Clustering. A.....n... ∀i = 1. K-MEANS(KM) • • • • • • • • • • • Read color image. Assign the pixels to the clusters as follows: ISSN : 0975-3397 Vol. V) = ∑ ∑ μ x −v (3) Where. v2…. where c is the number of clusters.. ∀i = 1. dij = ||xi − vj||. The value of m should be within the range m ∈ [1.. µij is the membership degree of pixel xi to the cluster centre vj and μij has to satisfy the following conditions: μ ∈ 0.. and approach binary functions. iii) Compute the fuzzy centres vj using: = ∑ ∑ ( ( ) ) (7) iv) Repeat step (ii) to (iii) until the minimum J value is achieved.... ∀i = 1.c. II. Segment the feature vector using FCM.. the membership functions become increasingly fuzzy. Enhance the image Convert the image into La*b* color space. the membership functions become more crisp..Manimala Singha et al. v2. The main aim is to minimize the objective function J(U.c (4) ∑ μ = 1 .

). Use MATLAB Functrion decorrstretch to perform the decorrelation. 3 No. The near infrared (NIR) band is important because of the high reflectance of chlorophyll in this part of the spectrum. The number of clusters were chosen. K-Means test1 Fuzzy C-Means test1 K-Means test2 Fuzzy C-Means test2 4.01 5. right red on the left edge. to segment the region of interest completely. have been implemented and tested on satellite images. as described in section IV.2a and 2b. The surface features have become much more clearly visible. / International Journal on Computer Science and Engineering (IJCSE) o • • A pixel x is assigned to cluster i such that the value of the membership function of x for i is maximum.30 ISSN : 0975-3397 Vol. To read multispectral bands from the LAN file using the MATLAB function multibandread. A 128-byte header is followed by the pixel values.lan. have been tested on La*b* color space and the results are shown in fig. The time taken to segment the satellite test image by different algorithms is given in Table 1.). III. in little-endian byte order [MATLAB 7. which enhances color separation across highly correlated channels. From the results. paris. in such a manner. They are stored as unsigned 8-bit integers. This is in conformity with the results of [24]. a large park on the western edge of Paris. The green area is the Bois de Boulogne. a large park (the Bois de Boulogne) located west of central Paris within a bend of the Seine River.65 4. EXPERIMENTAL RESULTS The algorithms are implemented on Matlab 7. green. red. The KM and FCM algorithms based on color segmentation. contains a 7-band 512x512 Landsat image (called test1).3]. Here we are reading band through multibandread are infrared. TABLE1. the proposed algorithms K-Means and Fuzzy C-Means.13 5. To enchance multispectral image is to use a decorrelation stretch.3 (The Mathworks Inc. Segmentation time of different algorithms on a test images: Algorithms Time in sec. for the image provided. The first test satellite images the LAN file. 2(a) and 2(b). The second satalite image downloaded from [25] of size 1024x1024 (called test2). It is even more useful when combined with visible red and green. for the same test satallite image. The spectral differences across the scene have been exaggerated.) for FCM method is more and the quality of the segmentation is much accurate than the KM method (fig. A property of color infrared composites is that they look red in areas with a high vegetation (chlorophyll) density. To segment the multispectral images. it can be conclude that the segmentation time (table 1. Create segment image. 12 Dec 2011 3759 .Manimala Singha et al. which are band interleaved by line (BIL) in order of increasing band number. Label every pixel in the image using the results from FCM.

12 Dec 2011 3760 .Manimala Singha et al. 3 No. / International Journal on Computer Science and Engineering (IJCSE) (a) (b) ISSN : 0975-3397 Vol.

for that we have used decorrelation stretch technique. The experimental result is shown in fig 2. IV. 12 Dec 2011 3761 . CONCLUSIONS Algorithms for segmentation of satellite images using Fuzzy C-Means (FCM). (c) K-means and (b). K-Means (KM) as described in section II. So the surface features have become much more clearly visible from the figures we can see that. have been implemented and tested on satellite image. Before doing image segmentation some lowresolution satellite images.Manimala Singha et al. Color image segmentation using (a). 3 No. it is often necessary to enhance multispectral image that is suitable for visual interpretation. (d) FCM. From the ISSN : 0975-3397 Vol. / International Journal on Computer Science and Engineering (IJCSE) (c) (d) Fig2.

2008.13.php [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] AUTHORS PROFILE Ms. 322–325. Man. Queensland. Fuzzy Sets and Fuzzy Logic-Theory and Applications. Bernardi. Ahn. Wu.” Image Coding using Wavelet Transform”. since 1998. degrees in Computer Science from Assam University.1 it seen that KM segmentation methods are fast as compared to FCM.14-22. “Fuzzy c-means clustering with spatial information for image segmentation”. M. vol.edu/database/database. September 2006. Stavros. 16th IEEE International Conference on Tools with Artificial Intelligence (ICTAI). M. Woods. Y. and J. and M. K. J. pp. as a Research Scholar and her area of interest includes image segmentation. 36. and Cybernetics. Bezdek. Laboratories Universities Bell. B. Pearson Education. T.9–15. 224-231. Plenum Press. Proceedings of international symposium on evolving fuzzy systems. 2000. Tzeng. G. Vol."Some methods of classification and analysis of multivariate observations". Issue 3. Degrees from Indian School of Mines. Y. 2006. Stuanton. L. Pattern recognition. 4. IEEE Transactions on Systems. Vol. Khan and A. 25(11): 1293-1302. M.1345-1354. Eddins. J. 35–67. “Digital Image Processing”.Sc. P. 3005-3011. 2004.7. M. Pattern recognition. L. G.D. H. M. T. J. B. of Computer Science. Elmaghraby. Dhanbad. T. ACM Computing Surveys. / International Journal on Computer Science and Engineering (IJCSE) Table. Silchar in 2005 and 2007 respectively. quantitative data and formulae”. J. Oct 2008. Kherfi.. Cairns. C. pp. and W. Q. Stiles. University of California Press. Antonini.1. E. April 2009. Lin. and S. E. Chien.S.D. N. 2. D. Barlaud. 28. no. 1. and Ph. P Mathieu. Jin and D. Kwon. 118-122. Signal Processing. 40. but segmentation methods using FCM give better segmented images with finer details and accurate location. “Fuzzy c-means clustering algorithm with a novel penalty term for image segmentation” Optoelectronic review.Manimala Singha et al. Assam University. J.Sc. and image searching in large databases. Kim. 28(1). Ehrlich. J. 2007. Gonzalez R. feature extraction. 1981. pp. ScienceEchoes 2006. pp. daubechies. A. 87. 255-258. 2005. Pearson Education. 30. K. S. Issue 4. Vol..‟Pattern Recognition and Fuzzy Objective Function Algorithms”. Proceedings of Fifth Berkeley Symposium on Mathematical Statistics and Probability. Silchar. 2000. International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS). W. M. Yang. “A modified fuzzy c-means image segmentation algorithm for use with uneven illumination patterns”. 309-315. Chen. Anirban and M. 1984. Woods. “Image segmentation by a fuzzy clustering algorithm using adaptive spatially constrained functions”. Chen. Yuan. R.Sc. Vol. 359-369. His areas of research interest are Image Processing. PHI. Computers and Geosciences 10. Li. McQueen. 12 Dec 2011 3762 . Hemachandran is associated with the Dept.C. S. D. J. pp. No. “Fast image segmentation based on K-Means clustering with histograms in HSV color space. J. A. 3 No. pp. 2nd edition. pp. Appl. Ahmad. Y. Kwak. No. ISSN : 0975-3397 Vol. 1967. 2nd ed. MacQueen. “A K-means clustering algorithm”. “A Switching vector median based on the CIELAB color space for color image restoration”.” Proceedings of IEEE International Workshop on Multimedia Signal Processing (MMSP2008). 395-405. Lotfy and A. pp. and A. “CoIRS: Cluster-oriented Image Retrieval System”. Martin. IEEE Transactions On Geoscience And Remote Sensing. May 1998. “Unsupervised Pixel Classification in Satellite Imagery Using Multiobjective Fuzzy Clustering Combined With SVM Classifier”. He obtained his M. E. vol. H. Zanaty. for her Ph. A. 2004. 2002. 2004. Wyszecki and W. Chen. Gonzalez and R. Ziou. http://sipi. Berkeley. REFERENCES [1] [2] [3] [4] G. Manimala Singha received her B. L.usc. pp. “FCM: The Fuzzy c-Means clustering algorithm”. A. 281–297. Li and R. C. H. Proc. “Color image segmentation using edge and adaptive threshold value based on the image characteristics”. New York. 191–203. C. 1979. 1967 pp. John Wiley and Sons. 2000.” Image Retrieval from the World Wide Web: Issues. Chuang. R. and P. and Systems” . Vol. IEEE transaction on image processing. Chen. pp. Stat. Hartigan and M. Tirupati and M. Wong. Proc. Y. Prof. “Image Segmentation Technique in Medical Sciences”. Ujjwal. Bezdek. vol. “Color science : concepts and methods.Tech. and S. 35. “Color image segmentation an innovative approach”. Pattern Recognition Letters. L. Vol. Tolias and M. Full. “Cluster center initialization algorithm for K-Means clustering”. Proceedings of 5th Berkeley Symposium on Mathematical Statistics and Probability. 2007. C. 47. J.pp-205-220. “Digital Image Processing using MATLAB”. April 1992. Degree from Sri Venkateswara University. Computerized Medical Imaging and Graphics. Y. Chen and Y. "Some Methods for classification and Analysis of Multivariate Observations". C. Zheng. Techniques. Australia. Lu. and I. and T.In. Klir and B. Software Engineering and Distributed Computing. S. Presently she is working. “A fuzzy clustering technique for medical image segmentation”. March 2004. R. pp. 100–108.

Sign up to vote on this title

UsefulNot useful- Face recognition System PPTby Bhushan Benz
- Segmentation by Fusion of Self-Adaptive SFCM Cluster in Multi-Color Space Componentsby AI Coordinator - CSC Journals
- Segmentation of Lung Glandular Cells using Multiple Color Spacesby Billy Bryan
- Survey on Clustering Based Color Image Segmentation and Novel Approaches to Fcm Algorithmby esatjournals

- Natural Scene Text Understanding
- IJFTET - Vol. 4-Issue 1_IMPROVED OTSU’S THRESHOLDING TECHNIQUES FOR HSI COLOR IMAGE SEGMENTATION
- A_New_Approach_Towards_Color_Based_Image_Segmentation-libre.pdf
- Image 1
- Face recognition System PPT
- Segmentation by Fusion of Self-Adaptive SFCM Cluster in Multi-Color Space Components
- Segmentation of Lung Glandular Cells using Multiple Color Spaces
- Survey on Clustering Based Color Image Segmentation and Novel Approaches to Fcm Algorithm
- Bx 4301429434
- Segmentation Tutorial
- ijcnn03_710
- International Journal of Engineering Research and Development (IJERD)
- Paper 12
- Color Based Segmentation
- Segmentation
- Collection (1)
- segmentation
- Color Segmentation Methods
- Image Segmentation guide
- 11.a Novel.full
- c 1104011322
- Color Image Segmentation Technique for Screen Printing
- remote.pdf
- IEEE_IP08
- Segmentation of Images by using Fuzzy k-means clustering with ACO
- NEURAL NETWORK BASED CLASSIFICATION AND DETECTION OF GLAUCOMA USING OPTIC DISC AND CUP FEATURES
- Klasifikasi Gambar Medis
- Image Proc_Term Paper Report
- IETECH
- Image Segmentation
- IJCSE11-03-12-085

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.

scribd