Globally convex variational model for multiphsae image segmentation

The study is to investigate a fast globally convex variational model for the multiphase image segmentation. Firstly, a nonconvex energy functional on the membership functions, which are used as indicators of different homogeneous regions, is introduced by incorporating edge-based information. Secondly, the nonconvex problem is converted into a continuous convex formulation. Finally, a dual minimization formulation of the binary partitioning function accurately describes disjoint regions using stable segmentations by avoiding local minima solutions and unambiguous segmentation. Experiments results show more accurate segmentation results on both medical and natural images compared with multi-region competition model.