Regionalized fuzzy C-means algorithm for segmentation of color remotesensing image

The improvement in spatial resolution of high resolution remote sensing images increases the internal spectral variability of each land cover class and induces geometric noise caused by tiny targets on it, so pixed based FCM and its improved algorithms can not overcome the difficulties during segmentation. Therefore, a regionalized fuzzy C-means (RFCM) algorithm is proposed for high resolution remote sensing image segmentation. By Voronoi tessellation, the image domain is partitioned into Voronoi polygons to fit the shapes of objects with the polygons. On the basis of the domain partition, the FCM's objective function is defined for the segmentation of high resolution remote sensing image. Test results show that the proposed algorithm is capable of segmenting high resolution remote sensing image with higher accuracy than the FCM and the enhanced FCM algorithm (EFCM). ©, 2015, Northeast University. All right reserved.