Polarimetric SAR image segmentation based on spatially constrained kernel fuzzy C-means clustering

A spatially constrained kernel fuzzy C-means (SCKFCM) algorithm is represented for polarimetric SAR (PolSAR) remote sensing image segmentation in this paper. Compared with classic fuzzy C-means (FCM) algorithm, kernel method could perform the nonlinear mapping from the original space to kernel space. Thus, SCKFCM is not impacted by the remote sensing image data distribution. Furthermore, in order to overcome the affection of speckle noises, the spatial constraint item is added in the objective function, which would improve the image segmentation accuracy effectively. The experiment results on PolSAR image segmentation demonstrate the validity of proposed SCKFCM approach.