Multiphase SAR Image Segmentation With $G^{0}$ -Statistical-Model-Based Active Contours

In this paper, we propose a variational multiphase segmentation framework for synthetic aperture radar (SAR) images based on the statistical model and active contour methods. The proposed method is inspired by the multiregion level set partition approaches but with two improvements. First, an energy functional which combines the region information and edge information is defined. The regional term is based on the G⁰ statistical model. The flexibility of G⁰ distribution makes the proposed approach to segment SAR images of various types. Second, we use fuzzy membership functions to represent the regions. The total variation of the membership functions is used to ensure the regularity. This not just guarantees the energy functional to be convex with respect to the membership functions but also enables us to adopt a fast iteration scheme to solve the minimization problem. The proposed method can segment SAR images of N regions with N - 1 membership functions. The flexibility of the proposed method is demonstrated by experiments on SAR images of different resolutions and scenes. The computational efficiency is also verified by comparing with the level-set-method-based SAR image segmentation approach.