The Split-Bregman-Projection Algorithm of a Variational Model for Multiphase Image Segmentation

摘要

In this paper, we propose a variational model for multiphase image segmentation by using n binary label functions to partition n regions. Our proposed model is subjected to n constraints due to the definition of binary functions and a constraint for avoidance of vacuum and overlapping. We use the convex relaxation approach to solve binary constraints of label functions. In order to satisfy the constraint of the second class, we develop the projection algorithm based on Lagrange multiplier method. To improve its computation efficiency, we design the Split Bregman algorithm for solution of every label function after convexification of it during alternating minimization. 2D and 3D multiphase image segmentation experiments are given to validate our model and the Split-Bregman-Projection algorithm.