SAR image segmentation using kernel density estimation on region adjacency graph

In this paper, we propose a new synthetic aperture radar (SAR) image segmentation scheme. Firstly, the SAR image is over-segmented using the mean shift (MS) algorithm while the original image discontinuity characteristics are preserved. Secondly, we propose a novel method to estimates the probability density function of each node on region adjacency graph (RAG) using kernel density estimation (KDE) method. This estimation includes the information of similarity and proximity of any pairs of nodes at the same time. Our approach yields superior performance and also is feasible for real-time processing.