基于GVF力的拓扑保持分割模型及其对偶算法  

Topology Preserving Segmentation Model Based on GVF Force and Its Dual Algorithm

作　　者：沈梦洁 潘振宽 宋金涛 魏伟波 SHEN Meng-jie;PAN Zhen-kuan;SONG Jin-tao;WEI Wei-bo(College of Computer Science & Technology, Qingdao University, Qingdao 266071, China)

机构地区：[1]青岛大学计算机科学技术学院,青岛266071

出　　处：《青岛大学学报（自然科学版）》2022年第1期1-10,18,共11页Journal of Qingdao University(Natural Science Edition)

基　　金：国家自然科学基金(批准号:61772294,11472144)资助;山东省联合基金(批准号:ZR2019LZH002)资助。

摘　　要：针对自排斥Snake模型对于狭窄图像区域作用力不足,传统加性算子分裂方法计算复杂,内存用量也会随着图像大小的增加而迅速增长等问题,提出了在原模型的基础上增加梯度矢量流有向力场,以加快轮廓线在图像狭窄区域的演化速度,并为改进的模型设计快速对偶算法以简化算法设计,提高求解效率。数值实验表明,改进模型及算法在计算效率方面较经典模型及算法有较大提高。In view of the insufficient force of the self-repelling Snake model on narrow image regions,the calculation of traditional additive operator splitting method is complicated,and the memory usage will also increase rapidly with the increase of the image size.It is proposed that the Gradient Vector Flow directed force field based on the original model was added to accelerate the evolution speed of the contour line in the narrow area of the image,and a fast Dual algorithm is designed for the improved model to simplify the algorithm design and improve the efficiency of the solution.Numerous numerical experiments show that the proposed improved model and algorithm have a greater improvement in computational efficiency than the classic model and algorithm.

关 键 词：自排斥Snake模型 拓扑保持分割 对偶算法 梯度矢量流 变分法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]