IOSR Journal of Electronics and Communication Engineering (IOSR-JECE) e-ISSN: 2278-2834,p- ISSN: 2278-8735.Volume 9, Issue 2, Ver. III (Mar - Apr. 2014), PP 67-73 www.iosrjournals.org www.iosrjournals.org 67 | Page Spatial clustering method for satellite image segmentation R.Ilanthiraiyan, V.M.Navaneetha Krishnan Department of Electronics and Communication Engineering, Dr.S.J.S.Paul Memorial College of Engg and Technology, Pondicherry University, Pondicherry, India. Abstract: spatial clustering method can be used for the analysis of satellite image to get the meaningful information with the help of fuzzy local information c-means algorithm using image segmentation technique. In previous clustering based image segmentation method may lose the important image details due to the noise present in the satellite image. The effects of noise are avoided by the spatial relationship among pixels, but it often generates boundary zones for the mix of pixel around the edges .To overcome these problem (FELICM) reduces the edge degradation by introducing the weights of pixel within the local neighboring windows. Canny edge detection used for the extraction of edges, during that detection multi-Otsu threshold can be used to obtain two adaptive thresholds. Local neighbors and window center are separated by edge with respect to different weights are set to the windows. Until the final clustering result may be obtained in a efficient manner the pixel of the different local neighbor window periodically iterated. Without any filter preprocessing steps FELICM method can directly applied to the satellite image to get the valuable information when compared to the remote sensing image of a experimental results and not only solves the problem of isolation of samples and random distribution of pixels inside the region but it also produce the high edge accuracies. Index terms: spatial clustering, image segmentation, canny edge detection, local information, multi-Otsu threshold. I. Introduction Isolation of object or samples, and ignores the spatial relationship with respect to the pixel image clustering is an effective method for image segmentation. Usually noise are present in that image, before clustering some smooth filters[1]-[3]are used to reduce the noise, and some other researcher to decrease the difference among the pixel[4],[5] in a region to use the texture description or spectral reduction as per the color clustering and color learning algorithm. The preprocessing steps of image clustering may lose some important image details and also it depends upon some important parameter like mean shift, and robustness of clustering. Number of algorithm[6]-[9] have proposed for clustering based image segmentation to make clustering are more robust such as bias corrected fuzzy c-means (FCM) (BCFCM) which can deal with the original image. The spectral features of pixel and mean filtered neighbors and the parameter of controls the effect of neighbors to determine the label of pixel in BCFCM method and also introduce the partial membership function for classifying the object or samples to different number of cluster present in that image using fuzzy factor. The combined results of support vector machine(SVM) and clustering using majority voting[6] are used to determine the homogeneous region in the hyper spectral (spectral –spatial) images and also ISODATA algorithm and Gaussian mixture resolving techniques used for the image clustering.[7] presented the new adaptive clustering algorithm is capable of utilizing local contextual information to impose the local spatial continuity that exploiting the inter-pixel correlation inherent in most of the real –world images. The proposed of Dulyakarn and Rangsanseri [8] spatial information with FCM improved the segmentation performance when compared to the remote sensing image results. FLICM can overcome the disadvantages of fuzzy c- means algorithm [10] and at the same enhance the clustering performance using the spectral and spatial information with the help of fuzzy factor, and it also noise insensitive method. In this method label of one pixel is related to label of its spatial neighbors, and the edges of each region will be dislocated due to the incorrect cluster of labels assigned to pixel. Therefore fuzzy c-means (FCM) with edge and local information (FELICM) reduces the edge degradation by introducing the weights of pixel within the local neighboring windows and also produce the high edge accuracies when compared to the FLICM. The implementation of spatial image clustering will be introduced in the section II. Experimental results and discussion will be addressed in the section III. Finally the conclusion and future enhancement work will be drawn.