计算机应用 ›› 2010, Vol. 30 ›› Issue (06): 1616-1618.

• 图形图像处理与模式识别 • 上一篇    下一篇

基于游程递归的连通区域标记算法

沈乔楠,安雪晖   

  1. 清华大学
  • 收稿日期:2009-12-30 修回日期:2010-02-22 发布日期:2010-06-01 出版日期:2010-06-01
  • 通讯作者: 沈乔楠

Connected component labeling algorithm based on run recursive method

  • Received:2009-12-30 Revised:2010-02-22 Online:2010-06-01 Published:2010-06-01
  • Contact: SHEN QiaoNan

摘要: 在研究已有算法的基础上,提出一种基于游程递归的标记算法,该算法可以对二值图像实现快速标记。顺序扫描图像,寻找未标记的游程,并递归搜索与之连通的游程,直到一个连通区域生成。在游程搜索过程中,在当前游程的相邻两行上,以其左端点为起始点分别向前向后进行连通游程的搜索;同时根据游程之间的位置关系对搜索策略进行优化,减少了重复搜索,提高了处理速度。该算法只需经过一次扫描图像,就能快速、准确地标记连通区域。在与已有算法的实验结果比较中,该算法具有较快的执行速度和较高的准确率,并且占用较少的内存,可以满足在施工现场中运动目标实时检测的需要。

关键词: 运动目标, 连通区域标记, 二值图像

Abstract: A run-based recursive labeling algorithm was proposed based on the previous algorithms for binary images. The algorithm scanned the whole image level by level until a run unlabeled was found and the runs connected to it were searched based on recursion until a connected component was generated. During the searching process, the connected runs forward and backward from the start point of current run on its neighboring levels. Then the searching method was optimized according to the relationship between the two runs and the search time was reduced without searching repeatedly. The proposed algorithm can complete the labeling with only one time of scanning and the comparative experiments with traditional algorithms show that it is more efficient and occupies less memory. It can real-time detect the moving targets on construction sites.

Key words: moving targets, connected component labeling, binary image