一种新的水平集图像分割模型  

New image segmentation model based on the level set method

作　　者：王卫卫[1] 杨塨鹏[1] 吕畅[1] 杨艳琦[1] 

机构地区：[1]西安电子科技大学理学院,陕西西安710071

出　　处：《西安电子科技大学学报》2013年第6期39-45,91,共8页Journal of Xidian University

基　　金：国家自然科学基金资助项目(60872138;61271294;61105011)

摘　　要：在Chan-Vese模型基础上,引入一个非凸的正则项,提出了一个新的变分水平集模型.一方面利用正则项的非凸性可起到更好的边缘保护作用,另一方面为水平集的演化增加了一个驱动力.同时,利用Nesterov算法实现了模型的快速求解.实验结果表明,与Chan-Vese模型相比,该模型在准确分割出图像目标的同时更好地刻画了边缘.A novel variational model based on the level set method is proposed. In the model, we introduce a nonconvex regularizer which performs better than the convex ones on protecting edges. On the other hand, the nonconvex regularizer is designed to provide a force which can help the evolution of the level set function. The Nesterov algorithm is introduced to implement the model. Experiments demonstrate the advantages of the proposed method over the Chan-Vese model in terms of both efficiency and accuracv.

关 键 词：图像分割 变分法 水平集 Nesterov算法 

分 类 号：TN911.73[电子电信—通信与信息系统]