1. ISSN: 2278 1323International Journal of Advanced Research in Computer Engineering & TechnologyVolume 1, Issue 4, June 2012Color Image Segmentationusing Clustering TechniquePatel Janak kumar Baldevbhai, R.S. Anand the literature, it is observed that different transforms are used AbstractThis work presents image segmentation technique to extract desired information from remote-sensing images orbased on colour features with K-means clustering algorithm. Inbiomedical images (Mehmet Nadir Kurnaz et al; 2005).this we did not used any training data. In this paper, we present Segmentation evaluation techniques can be generally divideda simple and efficient implementation of k-means clusteringalgorithm. The regions are grouped into a set of classes usinginto two categories (supervised and unsupervised). The firstK-means clustering algorithm. Results are grouped into clusters category is not applicable to remote sensing because anso avoiding feature calculation for every pixel in the image. optimum segmentation (ground truth segmentation) isAlthough the colour is not frequently used for imagedifficult to obtain. Moreover, available segmentationsegmentation, it gives a high discriminative power of regions evaluation techniques have not been thoroughly tested forpresent in the image. Here clusters are grouped & segmentationremotely sensed data. Therefore, for comparison purposes, itis obtained in form of colors through which important objectsare segmented, extracted or recognized.is possible to proceed with the classification process and thenindirectly assess the segmentation process through theIndex Termscolor Image segmentation, K-means, clusters,produced classification accuracies. (Ahmed Darwish, et al;unsupervised classification.2003).Clustering is a mathematical tool thatattempts todiscover structures or certain patterns in a data set, where theobjects inside each cluster show a certain degree ofI. INTRODUCTION similarity. he process of image segmentation is defined as: theTsearch for homogenous regions in an image and later theclassification of these regions. It also means the partitioning For image segment based classification, the images thatneed to be classified are segmented into manyhomogeneous areas with similar spectrum informationof an image into meaningful regions based on homogeneityfirstly, and the image segments features are extracted basedor heterogeneity criteria (Haralick et al; 1992). Image on the specific requirements of ground features classification.segmentation techniques can be differentiated into theThe colour homogeneity is based on the standard deviation offollowing basic concepts: pixel oriented, Contour-oriented, the spectral colours, while the shape homogeneity is based onregion-oriented, model- oriented, colour oriented and hybrid. the compactness and smoothness of shape. There are twoColour segmentation of image is a crucial operation in imageprinciples in the iteration of parameters:1) In addition toanalysis and in many computer vision, image interpretation, necessary fineness, we should choose a scale value as large asand pattern recognition system, with applications in scientific possible to distinguish different regions; 2) we should use theand industrial field(s) such as medicine, Remote Sensing, colour criterion where possible. Because the spectralMicroscopy, content- based image and video retrieval, information is the most important in imagery data, the qualitydocument analysis, industrial automation and quality controlof segmentation would be reduced in high weightiness of(Ricardo Dutra, et al;2008). The performance of colourshape criterion.segmentation may significantly affect the quality of an imageThis work presents a novel image segmentation based onunderstanding system (H.S.Chen et al; 2006).The mostcolour features from the images. In this we did not used anycommon features used in image segmentation includetraining data and the work is divided into two stages. Firsttexture, shape, grey level intensity, and colour. The enhancing color separation of satellite image using decorconstitution of the right data space is a common problem in relation stretching is carried out and then the regions areconnection with segmentation/classification. In order togrouped into a set of five classes using K-means clusteringconstruct realistic classifiers, the features that are sufficiently algorithm. Using this two-step process, it is possible torepresentative of the physical process must be searched. In reduce the computational cost avoiding feature calculationfor every pixel in the image. Although the colour is not Manuscript received June 19, 2012. frequently used for image segmentation, it gives a high Patel Janakkumar Baldevbhai is with the Image and Signal Processingdiscriminative power of regions present in the image.Lab., Electrical Engineering Department, Research Scholar, EED, IndianInstitute of Technology Roorkee, Uttarakhand, India on duty leave underColour segmentation is an essential issue with regard toQIP scheme of AICTE from the L.D.R.P. Institute of Technology & Research, vision applications, such as object detection and navigationGandhinagar, and Gujarat, India. (Corresponding author phone: (Bosch et al., 2007; Lin, 2007). The process of color09458121095; 079-23221371(R) e-mail: janakbpatel71@gmail.com). R.S. Anand is with the Electrical Engineering Department, Professor,segmentation consists of color representation, color featureEED, Indian Institute of Technology Roorkee, Uttarakhand, India extraction, similarity measurement and classification. In563All Rights Reserved 2012 IJARCET

2. ISSN: 2278 1323International Journal of Advanced Research in Computer Engineering & TechnologyVolume 1, Issue 4, June 2012color representation, the RGB (Red, Green and Blue) model,used to estimate the clustering index (Al Aghbari and Al-Haj,which expresses color as a mixture of red, green and blue 2006). The idea of a histon, which is an encrustation of athree color components, is often used to depict the color histogram such that the elements in the histon are the set of allinformation of an image (Bascle et al., 2007; Weng et al.,the pixels that can be classified as possibly belonging to the2007). By using a transformation, the secondary colors, same segment, was introduced for color segmentation bywhich are CMY (Cyan, Magenta and Yellow) or Murshrif and Ray (2008), and the total computation time thisRGGBBR, can be obtained and used as an alternative colorapproach requires for a 179X122 image is 2.41 s. Neuralmodel (Wang et al., 2007). The HSI model, which transformsnetworks (Bascle et al., 2007) have recently been used as aRGB into Hue, Saturation and Intensity, is also a popular clustering kernel for color segmentation, where componentscolor model at present, and its good performance has been of the RGB space and the intensity are used as inputs andshown in many works (Kim et al., 2007, 2008; Wangenheim three calibrated colour components are considered as outputset al., 2007). HSV (Value) and HSL (Luminance) are very of the modified multi-layer perceptron (MLP). After thesimilar to the HSI model due to the transformation formulas training procedure, good segmentation performance isapplied. Using the HSI color model, a specific color is able to achieved. Furthermore, the look-up tables (LUT) of thebe recognized regardless of variations in saturation andmodified MLP can be applied for real-time applications, sointensity. CIE Luv, CIE Lab and YCbCr (Wang and Huang,that the execution time for a 320X 240 image is only 0.003752006; He et al., 2007) are color spaces which represent a s. However, a huge database needs to be created for thiscolor by its lightness (L), luminance (Y) and chromaticitysystem to work, and if an input image is very different from(uv, ab and CbCr). The idea of color ratio was firstthose in the database, the network should be re-trained tointroduced by Barnard and Finlayson in 2000 to identify the improve the results. The well-known K-means methodshadow and non-shadow regions to be robust under(Lloyd) is one of the most commonly used techniques in thechanges in luminance. In 2002, the RGB ratio of the pixel clustering-based segmentation field for industrialvalue to the local sum (R/Rsum, G/Gsum, B/Bsum) was applications and machine learning (Berkhin, 2002; Mignotte,proposed by Finlayson et al. to deal with the influences of 2008). The fuzzy c-means theory (the fuzzy version ofshadows produced by variations in illumination. In addition,K-means) is applied as the clustering method (Kuo et al.,Finlayson et al. (2005) presented an alternative RGB ratio2008), and similarity measurement is based on Euclideandefinition, which is the ratio of the intensity of a pixel to the distance (Luis-Garcia et al., 2008). Bosch et al. (2007)local average (R/Rave, G/Gave, B/Bave), and this formula is presented an approach that can recognize grass, sky, snowused due to its invariance to luminance and device changes. and road using fuzzy logic with predefined classes, for whichIn this paper, we propose a new RGB ratio model, which is the average processing time for an image size of 180X120 tobased on the fact that a change in the intensity of a reference 250X250 is 60 s. Efficient fuzzy c-means clustering (qFCM)color will lead to a change in the RGB color components, butis also applied to speed up the clustering process by splittingtheir ratios to the reference color (R/Rref, G/Gref, B/Bref)a target image into several small sub-images (Chen et al.,will be linear to an intensity change (Benedek and Sziranyi,2005). The computation time that qFCM requires for a2007; Mikic et al., 2000). With this property, a specific color,128X128 gray-level image is 0.11.2 s. The use of a templatesuch as the reference Colour, can be described as a linearimage is another fast segmentation method. For instance, ancolor model, so that it is invariant to intensity variation.image database of eyes can be established, and a skin colourMoreover, information about the three color componentsdatabase can be obtained from a colour conversion matrix(RGB) is used to describe the chromaticity by the proposedwith color of the sclera. Consequently, fixed thresholds of theRGB ratio space. Therefore, while inheriting theHSV space are introduced to detect the skin area in an inputcharacteristics of HSI and RGB models, the RGB ratio hasimage (Do et al., 2007). However, the use of template imagesseveral advantages with regard object recognition under is restricted to specific objects, and may require a large imagevariations in intensity.database. In this paper, a dynamic fuzzy variable range isThere exist many complex and state-of-the-art techniques forproposed to achieve a high quality segmentation result.colour segmentation which are excellent at partitioning anFirstly, the linearity between the RGB ratio and intensity isinput image. For example, the global color statistics can beestimated by a linear progressive method and parameterrepresented by a set of overlapping regions and modeled by aestimation. Secondly, upper and lower boundaries aremixture of Gaussians (GMM), and a local mixture model isobtained statistically for each colour ratio. These boundariesdescribed by Markov Random Fields (Kato, 2008). Byare used to define the fuzzy membership functions ofcoloroptimizing parameters of the global and local models, the ratio clusters, which dynamically vary corresponding tomaximum likelihood is estimated and then a pixel can be intensity changes. The proposed fuzzy systems parameterclassified. Although this approach has good segmentationoptimization, undertaken using a back propagation neuralresults, a large number of iterations are necessary tonetwork, makes the fuzzy decision more adaptive and moredetermine the optimal parameters. As a result, 16 s ofeffective. Meijer (1992) used sine-wave sounds to transformcomputation time is needed for an image with a 256X256image information without any image pre-processing, while aresolution (Tai, 2007). Hill manipulation of the colour multi-resolution approach was introduced to image-to-soundhistogram is another widely used approach to achieve colour mapping by Capelle et al. (1998).segmentation. A three-dimensional histogram can be The present work is organized as follows: Section 2obtained by accumulating three colour components of pixels. describes the data resources and software used. Section 3Dominant hill detection and minor hill dismantling are then describes the enhancing colour separation of image using564 All Rights Reserved 2012 IJARCET 3. ISSN: 2278 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012decor relation stretching. Section 4 describes the K-means clusters to be located in the data. The algorithm thenclustering method. In section 5 the proposed method of arbitrarily seeds or locates, that number of cluster centers insegmentation of image based on colour with K-means multidimensional measurement space. Each pixel in theclustering is presented and discussed. Experimental resultsimage is then assigned to the cluster whose arbitrary meanobtained with suggested method are shown in section 6. vector is closest. The procedure continues until there is noFinally, section 7 concludes with some final remarks.significant change in the location of class mean vectors between successive iterations of the algorithms (Lille sand Mean shift-based clustering and Keiffer, 2000). As K-means approach is iterative, it is A clustering algorithm based on mean shift was proposed computationally intensive and hence applied only to image subareas rather than to full scenes and can be treated as in [13]. Unfortunately, it becomes impractical in the unsupervised training areas (Lillesand & Keiffer, 2000).context of texture segmentation due to the expensivecomputation required in order to find the nearest neighbours K-means-based clusteringof a point in a highdimensional space. Hence, in this work, an Due to its simplicity and good convergence properties, theapproximate version has been utilized. It starts by initializing iterative k-means algorithm is probably the most widely usedthe mean shift procedure on a given point and then iterates as clustering algorithm. However, it suffers from importantusual until a stationary point is reached. However, at eachdrawbacks, such as the requirement of specifying the numberiteration, all points involved in the mean shift computation of clusters and the non-deterministic results produced ifare marked as already visited. Therefore, they are not taken random initialization is used (which is often the case).as initial points anymore. These points are also assigned aIn order to overcome the aforementioned problems, avote regarding their membership to the cluster associatedwrapper for k-means, which is a variation of thewith the mode being detected. The algorithm repeats this resolution-driven clustering algorithm proposed in [11], hasprocedure with the remaining not visited points. been applied. It has two main stages: split and refinement. Once all mode candidates have been found, mode mergingRegarding the split stage, let us assume that the data pointsis performed by means of the same approximate mean shift have been split intoalgorithm by considering the found modes as data points. IfC disjoint clusters (initially C = 1). The mean distancetwo modes are merged, their membership votes are alsobetween the centroid and its associated points (intra-clustermerged, thus keeping track of the new cluster structure. The mean distance) is computed for each cluster and the globalmode merging step is repeated until no modes are merged. mean distance (mean of intra-cluster mean distances) is obtained for the whole partition. If this global mean distanceMembership of each point is finally determined by majority exceeds a threshold, the largest cluster in terms ofvoting. intra-cluster mean distance is split into two. The split is done by finding the main principal component of the cluster andGraph clustering based on the normalized cut initializing two new child centroids at c d, where c is the centroid of the cluster to be split and d = 2/, with being The graph clustering algorithm based on the normalizedthe eigenvalue associated with the main principal componentcut proposed in [14] has become popular in the last years. . After the split stage, the refinement stage consists ofHowever, the main drawback of this approach is that theapplying k-means using the (C + 1) available centroids ascomputational technique for minimizing the normalized cutinitial seeds. Both split and refinement are iterated until nois based on eigenvectors. Thus, it suffers from scalabilitynew clusters are generated.problems, since in cases where the number of data points is The proposed wrapper has two main advantages over thevery large, eigenvector computation becomes prohibitive. classical k-means. First, instead of the desired number ofRecently, Dhillon et al. [15] proposed a more efficientclusters, the mean distance threshold controls the output oftechnique referred to as GRACLUS, which embeds a the algorithm.Such a threshold is more intuitive and closelyweighted kernel k-means algorithm into a multilevelrelated to perceptual properties than the number of clusters.approach in order to optimize locally the normalized cut.Second, the algorithm always behaves in the same way given However, before applying GRACLUS to the pattern the same input. Therefore, there is no need for runningdiscovery stage, the problem of specifying the number of different trials and keeping the best set of clusters accordingclusters must be addressed such as with k-means. Usually,to some criterion as it is the case when the initialization stepthe alternative is to first bipartition the whole graph and then of k-means has a random component.repartitions the already segmented parts if the normalized cutis below a specified value [14]. Colour-Based Segmentation Using K-Means Clustering ThebasicaimistosegmentcolorsinanautomatedfashionusingthII. K-MEANS CLUSTERING eL*a*b*colorspaceandK-meansThere are many methods of clustering developed for a wideclustering.Theentireprocesscanbesummarizedinfollowingstevariety of purposes. Clustering algorithms used forps.unsupervised classification of remote sensing image data Step1:Readtheimagevary according to the efficiency with which clustering takesReadtheimagefrommother source whichisin.JPEGformat.place (John R Jenson, 1986).K-means is the clusteringStep2:ForcolorseparationofanimageapplytheDecoralgorithm used to determine the natural spectral groupings relationstretching.present in a data set. This accepts from analyst the number of Step3:ConvertImagefromRGBColorSpacetoL*a*b*ColorSpace. 565All Rights Reserved 2012 IJARCET 4. ISSN: 2278 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012 Howmanycolorsdoweseeintheimage ifweignorevariations unsupervised problem into a supervised one.inbrightness?Therearethree colors:white,blue,andpink.As its name suggests, a pixel-based classifier aims atWecaneasilyvisuallydistinguish thesecolorsfromoneanother.determining the class to which every pixel of an input imageTheL*a*b*colorspace(alsoknownasCIELABbelongs, which leads to the segmentation of the image as aorCIEL*a*b*)enablesustoquantifythese visualdifferences. Thecollateral effect.L*a*b*colorspaceisderivedfromtheCIEXYZtristimulusvalues. In order to achieve this objective, several measures are computed for each image pixel by applying a number ofThe texture feature extraction methods as described in Section 3.1.L*a*b*spaceconsistsofaluminositylayerL*,chromaticity-layera*indicatingwherecolorfallsalongthered-greenaxis,and Classification with multiple evaluation window sizeschromaticity-layerb*indicatingwherethecolorfallsalongthe Although previous works on supervised pixel-basedblue-yellow axis.Allofthecolorinformation classification have already shown the benefits of utilizingisinthea*andb*layers.Wecanmeasurethedifference multiple evaluation window sizes [10, 11], which approach isbetweentwocolorsusingtheEuclideandistancemetric.Convertthe the best for combining these different sources of information isimagetoL*a*b* colorspace.still an open issue.Step4:ClassifytheColorsina*b*SpaceUsingK-MeansClustering For instance, in [10], different window sizes were integrated.by assigning a weight to their corresponding probabilities Clusteringisa way according to how well each window size separates a giventoseparategroupsofobjects.K-meansclusteringtreatseachtraining pattern from the others. However, since the trainingobjectashavingalocationinspace. Itfindspartitionspatterns are single-textured images, the assigned weight is notsuchthatobjectswithineachclusterareasclosetoeach representative of the structure of the test image, which in turn is composed of multiple texture patterns. Furthermore, thisotheraspossible,andas farfromobjectsinotherclustersas method may be biased to the largest window, as it capturespossible.K-meansclusteringrequires more information and, hence, has better capabilities ofthatyouspecifythenumberofclusterstobepartitioned distinguishing between texture classes. Later, in [11],andadistancemetrictoquantifyhowimproved classification rates were obtained by directly fusingclosetwoobjectsaretoeachother.Sincethecolorinformation the outcome of multiple evaluation window sizes using theexistsinthea*b*space,yourKNN rule. The main problem with this approach is that it doesobjectsarepixelswitha*andb*values. UseK-meanstocluster not guarantee that the most appropriate window size willtheobjectsintothreeclusters usingtheEuclideandistancemetric. always receive the majority of votes.Step5:LabelEveryPixelinthe Ideally, the strategy for classifying a test image usingImageUsingtheResultsfromK-MEANSmultiple evaluation window sizes should apply large windows Foreveryobjectinourinput,K-meansreturnsanindexcorrespon inside regions of homogeneous texture in order to avoid noisy classified pixels and small windows near the boundariesding toacluster.Labelevery pixelin between those regions in order to define them precisely.theimagewithitsclusterindex. Unfortunately, that kind of knowledge about the structure ofStep6:CreateImagesthatSegmenttheImagebyColor.the image is only available after it has been segmented. Usingpixellabels,wehavetoseparateobjectsinimagebycolor, Notwithstanding, an a priori approximation of that strategy canwhichwillresultinfiveimages. be devised through the following steps:Step 7: Segment the Nuclei into a Separate Image Step 1: Select the largest available evaluation window andThen programmatically determine the index of the cluster classify the test image pixels labelled as unknown (initially, allcontaining the blue objects because K means will not return thepixels are labelled as unknown).same cluster idx value every time. We can do this using theStep 2: In the classified image, locate the pixels that belongcluster center value, which contains the mean a* and b*to boundaries between regions of different textures and markvalue for each cluster.them as unknown, as well as their neighbourhoods. The size of the neighbourhood corresponds to the size of the 1. Select k -seeds s.t. d ( ki , k j ) > d minwindow used to classify the image. Step 3: Discard the current evaluation window. 2. Assign points to clusters by min dist. Step4: Repeat steps 1 to 3 until the smallest evaluationCluster ( pi ) = Arg min ( d ( pi , s j )) window has been utilized. This scheme, which can be thought of as a top-downs j { s1 ,, sk } approach, has been used during the supervised classification 3. Compute new cluster centroids: stage of the proposed segmentation technique. In addition to closely approximating the previously described ideal strategyCj 1 pi for using multiple evaluation window sizes, this approach avoids the classification of every image pixel with all the n pi jthcluster available windows. Hence, it leads to a lower computation 4. Reassign points to clusters (as in 2 above)time than previous approaches. 5. Iterate until no points change clustersSupervised pixel-based classification III. RESULTS AND DISCUSSIONAt this stage, the set of texture patterns found by the previousstage are used as texture models for a supervised pixel basedWe implemented proposed algorithm and tested itsclassifier, thus effectively transforming the original performance on a number of standard images of Mat Lab566 All Rights Reserved 2012 IJARCET 5. ISSN: 2278 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012software. We have used Peppers, Planet, Lena images fromMat Lab software as a standard image. Addition to theseimages we have implemented this proposed algorithm onheart image also & obtain segmentation results. Figure 1(a)shows original image of Peppers.png image and figure1(b)-1(g) show various segmented objects from originalimage. Here various color clusters and segmented objects areclearly visible. Table 1 shows parameter values ofPeppers.png image like Min, Max, mean, median, mode, Standard Deviation and range. Figure 1(h) showsthe scatter plot of original image Peppers.png. Figure 1 (i)shows Scatter plot with Bar and values of Peppers.png image.Figure 1 (j) shows Graph of Parameter values of Peppers.pngimage. Figure 1 (k) shows Radar Graph of Parameter valuesof Peppers.png image. Figure 2 (a) shows the second imageof our test data image of original Planets standard image frommat lab software. Figure 2(b) and 2(c) shows ObjectFigure 1 (c) Object Segmentation from Peppers image havingSegmentation from Planets image. Table 2 shows Parameter light green colorValues of Planets image. Figure 2(d) shows Scatter plot ofPlanets image. Figure 2 (e) represents Graph of Parametervalues of Planets image and Figure 2 (f) represents RadarGraph of Parameter values of Planets image. Similarly Figure3 shows results for Lena Image. Figure 4 shows segmentationresults of Heart image. Figure 1 (d) Object Segmentation from Peppers image having red colorFigure 1 (a) Original Peppers standard image from matlab Figure 1 (e) Object Segmentation from Peppers imageFigure 1 (b) Object Segmentation from Peppers image havingorange color 567All Rights Reserved 2012 IJARCET 6. ISSN: 2278 1323International Journal of Advanced Research in Computer Engineering & TechnologyVolume 1, Issue 4, June 2012 22060 20050 180Black40 160105 x min150 x max127 x mean30 140 128 x median 136 x mode 1209.173x std20RedGreen 100Violet10MagentaYellow80 100120140160180200 220Figure 1 (f) Object Segmentation from Peppers imageFigure 1 (i) Scatter plot with Bar and values of Peppers.png image Figure 1 (j) Graph of Parameter values of Peppers.png imageFigure 1 (g) Object Segmentation from Peppers imageBlack X Min Scatterplot of the segmented pixels in a*b* space Yellow Y250 Black Y 220200 MaxYellow X150Red X 200100 mean Magenta 500Red Ymedian 180YMagenta mode Green XX b* values 160stdViolet Y Green Y 140Violet Xrange 120 100 80Figure 1 (k) Radar Graph of Parameter values of Peppers.png100 120 140 160180200220 a* values imageFigure 1 (h) Scatter plot of Peppers.png image 568 All Rights Reserved 2012 IJARCET 7. ISSN: 2278 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012Table 1: Parameter Values of Peppers.png imagePeppers.pngMin MaxmeanmedmodeSTD ran geBlack X105 150127.21128136 9.172945Black Y126 160147.13148148 6.684334Red X106 156122.8 120115 9.467 50Red Y152 176165 165167 4.936 24Green X156 201183.05185187 7.955945Green Y133 201169.10168173 12.5043 68Violet X 128 179155.5 156168 12.93 51Violet Y 176 214202.4 204204 7.818 38Magenta X110 156126.3 123121 9.347 37Magenta Y163 200181.3 182185 6.347 37Yellow X 126 184147.6 147147 4.6658Yellow Y 92153115.5 115115 6.838 61Figure 2(c) Object Segmentation from Planets image Scatterplot of the segmented pixels in a*b* space200180160b* values1401201008060 120 130 140 150160170 180 190200Figure 2 (a) Original Planets standard image from matlab a* values Figure 2(d) Scatter plot of Planets imageTable 2: Parameter Values of Planets image250 Planets.jpgMinMax meanmed modestd range Red X120161 134.2 133 132 4.6 41 200 Red Y61 121 97.84 969511.52 60 Violet X 120199 134.7 131 128 12.01 79150 Red X Violet Y 118192 130.8 127 128 12.174Red Y100Violet X 50Violet Y0 Figure 2 (e) Graph of Parameter values of Planets imageFigure 2(b) Object Segmentation from Planets image 569All Rights Reserved 2012 IJARCET 8. ISSN: 2278 1323International Journal of Advanced Research in Computer Engineering & TechnologyVolume 1, Issue 4, June 2012 Min 200 range 150Max 100 Red X50 Red Y 0 std meanViolet X Violet Y modemedian Figure 2 (f) Radar Graph of Parameter values of PlanetsimageFigure 3(b) Object Segmentation from Lena image Figure 3(c) Object Segmentation from Lena imageFigure 3 (a) Original standard image of Lena from matlab 570 All Rights Reserved 2012 IJARCET 9. ISSN: 2278 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012 Lena.tif MinMa mea mediamod std rang f xn ne e Black X168190173.7 174174 2.9122 3 Black Y140187151.3 151149 3.9947 1 Red X166182171 171172 2.4016 2 Red Y127148142 142143 3.2921 5 Green X147176161 163165 5.9629 2 Green Y124143133.6 134141 5.3219 2 Violet X 132178161.1 162163 4.5146 9 Violet Y 90 125116.2 117120 5.7135 4 Magenta125148139.5 139138 3.6323Figure 3(d) Object Segmentation from Lena imageX Magenta109182143.4 141139 8.5273 Y 3 Yellow 133169157.9 158156 6.0836 X 1 Yellow 142210152.1 151146 6.8568 Y 5 Table 3: Parameter Values of Lena image Scatterplot of the segmented pixels in a*b* space220200180b* values160140120100Figure 3(e) Object Segmentation from Lena image80 120 130140150 160 170 180190 a* values Figure 3(f) Scatter plot of Lena image 571All Rights Reserved 2012 IJARCET 10. ISSN: 2278 1323International Journal of Advanced Research in Computer Engineering & TechnologyVolume 1, Issue 4, June 2012250200 Min150Max mean100 median 50mode stdFigure 4 (b) Segmented object1 of Heart image0 range Magenta X Magenta Y Violet XViolet YYellow Y Red YYellow X Red X Green Y Green X Black Y Black XFigure 3(g) Graph of Lena image parameter valuesFigure 4 (c) Segmented object2 of Heart imageMin Black XYellow 250 Y 200Black Y MaxYellow X 150 Red Xmean 100 Magenta50 0 Red YmedianYMagenta mode Green XXViolet YGreen Y std Violet XrangeFigure 4 (d) Segmented object3 of Heart imageScatterplot of the segmented pixels in a*b* space200180Figure 3(h) Radar Graph plot of Lena image parametervalues160b* values140120100 80 60110 120 130140 150 160170180190 200a* valuesFigure 4 (e) Scatter plot of Heart imageFigure 4 (a) Original image of Heart 572All Rights Reserved 2012 IJARCET 11. ISSN: 2278 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012Figure 5 Quantitative Comparison of Segmentation Methods Table 4: Methods573All Rights Reserved 2012 IJARCET 12. ISSN: 2278 1323 International Journal of Advanced Research in Computer Engineering & Technology Volume 1, Issue 4, June 2012 [4]Jean-Christophe Devaux et al; Aerial colour image segmentation byKarhunen-Loeve transform, 0-7695-0750-6, IEEE 2000, pp 309-312. [5]Jun Tang, A color image segmentation algorithm based on regiongrowing, 978-1-4244-6349-7, IEEE, vol 6, 2010, pp. 634-637 [6]Mehmet Nadir Kurnaz,et al; Segmentation of remote-sensing imagesby incremental neural network, Pattern Recognition Letters 26(2005) 10961104, pp 1096-1103. [7]N Bartneck et al; Colour segmentation with polynomialclassification,0-8186-2915-0/92, 1992, pp. 635-638. [8]Nae-Joung Kwak et al; color image segmentation using edge andadaptive threshold value based on the image characteristics,IEEEproceeding0-7803-8639-6,2004, pp 555-558. [9]Lingkui Meng, et al, Study on Image Segment Based Land UseClassification and Mapping, 2009 IEEE, pp [10] Pal N.R.et al; A review on image segmentation techniques, PatternRecognition 26(9), 1993, pp1277-1294. [11] Ricardo Dutra da Silva et al; Satellite image segmentation usingwavelet transforms based on color and texture features, ISVC 2008,part II, LNCS 5359, 2008, pp 113-122 [12] Robert A Schowengerdt, Remote sensing- models and Methods forImage Processing, IIIrd edition, Elsevier Inc. [13] T. W. Chen, Y. L. Chen and S. Y. Chien. Fast image segmentationbased on K-Means clustering with histograms in HSV color space. InProceedings of 10th IEEE Workshop on Multimedia SignalProcessing, 2008. [14] C.W. Chen, J. Luo, K.J. Parker, Image segmentation via adaptiveK-mean clustering and knowledge based morphological operationswith biomedical applications, IEEE Transactions on ImageProcessing, Vol.7 (12), 1998, pp 1673-1683. [15] B. Sowmya, B. Sheelarani, Colour Image Segmentation Using SoftComputing Techniques.International Journal of Soft Computing Applications, 4:69-80, 2009. [16] M. Mirmehdi, M. Petrou, Segmentation of color textures, IEEE Trans.Pattern Anal. 22 (2000) 142-159. [17] S.C. Kim, T.J. Kang, Texture classification and segmentation usingwavelet packet frame and Gaussian mixture model, Pattern Recogn. 40(2007) 1207-1221. [18]D. Puig, M.A. Garcia, Automatic texture feature selection for imagepixel Classification, Pattern Recogn. 39 (2006) 1996-2009. [19] J. Melendez, M.A. Garcia, D. Puig, Efficient distance-basedper-pixeltexture classification with Gabor wavelet filters, PatternAnal. Appl. 11(2008) 365-372. [20] M. Omran, A. Engelbrecht, A. Salman, An overview of clusteringmethods, Intell. Data Anal. 11 (2007) 583-605. [21]D. Comaniciu, P. Meer, Mean shift: A robust approach toward featureFigure 6 Quantitative Comparisons of Segmentation Space analysis, IEEE Trans. Pattern Anal. 24 (2002) 603-619.Methods[22]J. Shi, J. Malik, Normalized cuts and image segmentation, IEEETrans.Pattern Anal. 22 (2000) 888-905.IV. CONCLUSION [23] I.S. Dhillon, Y. Guan, B. Kulis, Weighted graph cuts withouteigenvectors: A multilevel approach, IEEE Trans. Pattern Anal. 29(2007) 1944-1957.We have presented an efficient implementation of k-means [24]D. Tsujinishi, Y. Koshiba, S. Abe, Why pairwise is better thanclustering algorithm. The algorithm has been implementedoneagainst-all or all-at-once, in: Proceedings of the IEEE IJCNN,on standard images from mat lab software. Results are plotted 2004, pp.693-698.in scatter plots showing the clusters & Radar plot showing [25] W.-Y. Ma, B.S. Manjunath, Edge Flow: A technique for boundarydetection and image segmentation, IEEE Trans. Image Process. 9the data analysis of clusters. Various segmentation methods (2000) 1375-1388.are given in form of chart. The plot of segmentation method[26] A.Y. Yang et al., Unsupervised segmentation of natural images viashows unsupervised k means cluster Method is better aslossydata compression, Comput. Vis. Image Und. 110 (2008) 212-225.compared to supervised classification segmentation methods.Janak B. Patel (born in 1971) received B.E.And the more well separated the clusters, the faster the(Electronics & Communication Engg from L.D.algorithm runs. This algorithm is significantly more efficientCollege of Engg. Ahmedabad, and M.E.than the other methods. (Electronics Communication & System Engg.) in2000 from DDIT. He is Asst. Prof. & H.O.D. at REFERENCES L.D.R.P.I.T.R., Gujarat. He is pursuing Ph.D. atIndian Institute of Technology, Roorkee.[1] Ahmed Darwish, et al, Image Segmentation for the Purpose OfObject-Based Classification,, 2003 IEEE pp. 2039-2041RR.S. Anand received B.E., M.E. and Ph.D. in[2] Darren MacDonald, et al; Evaluation of colour image segmentationElectrical Engg. from University of Roorkee inhierarchies, proceeding of the 3rd Canadian conference on 1985, 1987 and 1992, respectively. He is aComputer and Robot Vision, IEEE, 2006.professor at Indian Institute of Technology,[3] H C Chen et al, Visible color difference-based quantitative evaluationRoorkee. He has published more than 100of colour segmentation, IEEE proceeding, Vis image signal process research papers in the area of image processing andvol.153 No.5 Oct 2006 pp 598-609. signal processing.574 All Rights Reserved 2012 IJARCET