Fuzzy Clustering of Multispectral Images Based on Morphological Operators and Gustafson-Kessel Clustering

The noise reduces certitude in the results obtained of processing on the multispectral remote sensing images. Hence, the clustering of these images is very difficult due to noise. In this paper, a fuzzy clustering technique for this images that the morphological operators and median filter is used to reduce noise is offered. In this approach, at first, opening and closing morphological operators are used until tiny points that are similar noise and small holes in the image are exist. Thus, multispectral image is improved and boundary betweendifferent regions of image is preserved. Then, the bandsobtained of this operations algorithm are clustered by FCMclustering and the membership matrix obtained by using Gustafsson-Kessel clustering algorithm to detect clusters withdifferent shapes be improved. Finally, using the median filter, noise remains in the clustered image by selection anappropriate window is reduced. The proposed clustering algorithm has been done on the FLC1 multispectral image that has 12 bands. The results of experiment the proposed algorithm and compare with conventional fuzzy clustering methods, shows the feasibility and efficiency the proposed algorithm

کلیدواژه ها:

Multispectral Images ، Mathematical Morphology ، Opening Operator ، Closing Operator ، Fuzzy C-Means Clustering ، GK Clustering ، Median Filter