Home Articles Satellite Image Segmentation Based on Fuzzy C-Means Clustering

Satellite Image Segmentation Based on Fuzzy C-Means Clustering


Satellite Image Segmentation Based on Fuzzy C-Means Clustering

Fateme Ameri

KNT University of technology,
Email: [email protected] com

Mohammad Javad Valadan Zoej
KNT University of technology
Email: [email protected]

Mehdi Mokhtarzade
PhD student of photogrammetry
KNT University of technology
Email: [email protected] yahoo.com

There are several methods for segmenting images based on two fundamental properties of the pixel values: One of them is “discontinuity” that uses the discontinuities between gray-level regions to detect isolated points, edges and contours within an image. The other is “similarity” that uses decision criteria to separate an image into different group based on the similarity of the pixel levels. Clustering is one of the methods of second category. Clustering algorithms attempt to separate a dataset into distinct regions of membership. C-Means clustering is one of them. Furthermore, mixed pixels (mixels) in the image, which are not completely occupied by a single and homogeneous object, occur because the pixel size may not be fine enough to capture details on the ground. Fuzziness occurs due to the presence of mixels and use of fuzzy methods makes the results more reliable. Fuzzy methods in remote sensing have received growing interest for their importance in situations where the geographical phenomena are inherently fuzzy. Integration of these two techniques (C-Means clustering & Fuzzy methods) leads to Fuzzy C-Means clustering (FCM) that consider each cluster as a fuzzy set. Computational steps of FCM algorithm are: choosing the number of classes and the initial value for the means, classify the image by defining membership value for each class and assigning the pixels to the class corresponding to the closest mean, Re-computing the means of the class and at last, if the change in any of the means is more than some pre-defined small positive value, then stopping, else reclassifying the image based on membership functions and iterating the algorithm. This paper attempts to segment IKONOS satellite image based on FCM algorithm and detect different road classes on it .Furthemore result of this algorithm will be compare with what has been implemented by FCM algorithm of PCI Geomatica9.1 software