Multispectral images segmentation using new fuzzy cluster centroid modified

The presence of outliers, noise, corrupt pieces of data and great quantity of samples in a multispectral image, makes the segmentation analysis work tedious. The fuzzy clustering approach, specially, is susceptible to inhomogeneity of characteristics. Furthermore, many algorithms such us FCM, PFCM, FCC, FWCM and modification aim to solve these problems by integrating spacial information. This process is carried through the analysis of the sample's neighborhood. This paper proposes the integration of the sample presence probability into a ”term” like form inside the existent model NFCC. This algorithm presents the basic steps for fuzzy clustering. With a middle variant that integrates the measure between each sample to all the centroids, this replaces the existent term by a new term. This new term integrates the spatial relationship between each sample of the multispectral image into a fitting term. The method is applied to multispectral images. Overall accuracy indicates that the term integrated to NFCC model decrease the overall cluster overlapping.