Computer Engineering and Applications ›› 2013, Vol. 49 ›› Issue (3): 226-229.

Previous Articles     Next Articles

BandCut: interactive fast object segmentation using graph cuts

WANG Yarong, XU Qiuping   

  1. Training Department, Engineering College of Armed Police Force, Xi’an 710086, China
  • Online:2013-02-01 Published:2013-02-18

基于窄带图割的交互式快速目标提取

王亚荣,徐秋平   

  1. 武警工程学院 训练部,西安 710086

Abstract: Aiming at the limitations of critical point problems in local minima trap and low extracting efficiency, by means of graph cuts theory, an object extracting algorithm, BandCut, is proposed. An annular band region that encompasses object boundary is obtained in an interactive way. The distance map and s-t network are created successively. The object boundary is extracted via min-cut of s-t network. The experimental results show optimal and rapid extraction ability.

Key words: object extraction, color image segmentation, graph cuts, combinatorial optimization

摘要: 针对传统算法易陷入局部极值、提取效率不高的不足,运用图割理论,提出一种将目标提取问题转化为能量最小化的组合优化问题的BandCut算法。BandCut通过人机交互获取一个将目标边界包围在内的环状窄带区域,对该区域生成距离图,构造s-t网络,进行最小代价切割获取目标。实验表明,BandCut能获取最优解,提取效率是GrabCut的5倍。

关键词: 目标提取, 彩色图像分割, 图割, 组合优化