基于变宽邻域图割和活动轮廓的目标分割方法  

Object Segmentation Method Based on Variable Contour Neighborhood Graph Cuts and Active Contours

作　　者：徐秋平[1,2] 郭敏[1] 

机构地区：[1]陕西师范大学计算机科学学院,西安710062 [2]武警工程学院教育技术中心,西安710086

出　　处：《计算机工程》2009年第8期233-235,238,共4页Computer Engineering

基　　金：陕西省自然科学基金资助项目(2005A12);陕西师范大学研究生培养创新基金资助项目(2008CXS025)

摘　　要：基于图割的活动轮廓算法是一个结合图割优化工具和活动轮廓模型迭代变形思想的目标分割算法。针对算法在迭代过程中对已达目标边界的活动轮廓线所在邻域重复切割的不足,将活动轮廓线分为已达目标曲线段和未达目标曲线段,仅对未达目标曲线段进行膨胀得到可变宽度轮廓线邻域,从而减少对邻域的切割时间。实验表明,改进算法效率提高为原来的2~3倍。Graph Cuts Based Active Contours（GCBAC） approach is a combination of the optimization tool of graph cuts and the iterative deformation idea of active contours. An improved algorithm based on Variable Contour Neighborhood（VCN） is proposed to solve the disadvantage that GCBAC repeatedly cuts the contour neighborhood even these parts of the contour have reached the object boundary. The active contour is classified into the boundary-reached part and the boundary-unreached one, and the contour neighborhood is only dilated from the boundary-unreached one, so it sharply decreases the graph-cut time. Experimental results show that the efficiency of the improved algorithm is 2-3 times as high as the original one.

关 键 词：目标分割 活动轮廓 图割 组合优化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]