A Coastal Zone Segmentation Variational Model and Its Accelerated ADMM Method  

作　　者：HUANG Baoxiang CHEN Ge ZHANG Xiaolei YANG Huan 

机构地区：[1]College of Computer Science and Technology,Postdoctoral stations of system science,Qingdao University [2]College of Information Science and Engineering,Ocean University of China

出　　处：《Journal of Ocean University of China》2017年第6期1081-1089,共9页中国海洋大学学报（英文版）

基　　金：supported by the China Postdoctoral Science Foundation under Grant No.2015M571993;the Shandong Provincial Natural Science Foundation,China under Grant No.ZR2017MD004;the National Natural Science Foundation of China under Grant No.61602269;Qingdao Postdoctoral Application Research Funded Project

摘　　要：Effective and efficient SAR image segmentation has a significant role in coastal zone interpretation. In this paper, a coastal zone segmentation model is proposed based on Potts model. By introducing edge self-adaption parameter and modifying noisy data term, the proposed variational model provides a good solution for the coastal zone SAR image with common characteristics of inherent speckle noise and complicated geometrical details. However, the proposed model is difficult to solve due to to its nonlinear, non-convex and non-smooth characteristics. Followed by curve evolution theory and operator splitting method, the minimization problem is reformulated as a constrained minimization problem. A fast alternating minimization iterative scheme is designed to implement coastal zone segmentation. Finally, various two-stage and multiphase experimental results illustrate the advantage of the proposed segmentation model, and indicate the high computation efficiency of designed numerical approximation algorithm.Effective and efficient SAR image segmentation has a significant role in coastal zone interpretation. In this paper, a coastal zone segmentation model is proposed based on Potts model. By introducing edge self-adaption parameter and modifying noisy data term, the proposed variational model provides a good solution for the coastal zone SAR image with common characteristics of inherent speckle noise and complicated geometrical details. However, the proposed model is difficult to solve due to to its nonlinear, non-convex and non-smooth characteristics. Followed by curve evolution theory and operator splitting method, the minimization problem is reformulated as a constrained minimization problem. A fast alternating minimization iterative scheme is designed to implement coastal zone segmentation. Finally, various two-stage and multiphase experimental results illustrate the advantage of the proposed segmentation model, and indicate the high computation efficiency of designed numerical approximation algorithm.

关 键 词：coastal zone SEGMENTATION VARIATIONAL POTTS model ALTERNATING direction method with MULTIPLIERS edge self-adaption 

分 类 号：P748[天文地球—海洋科学]