一种新的边界跟踪算法.PDFVIP

2011 年 工 程 图 学 学 报 2011 第3 期 JOURNAL OF ENGINEERING GRAPHICS No.3 一种新的边界跟踪算法 石 爽， 曲仕茹， 何 力 （西北工业大学自动化学院，陕西 西安 710072 ） 摘 要：针对提取的图像边缘中存在非单像素和断点的情况，提出了双层边界区域 生长的边界跟踪算法。通过对中心点周围里层点和外层点分别进行搜索，然后把里层点和上 一层中心点的外层点合并，并将并集中的点分别作为下一步搜索的中心点，循环向下搜索。 同时充分考虑了起始中心点单向搜索的情况，并在一次搜索过程中完成了对断点的补齐工 作，从而弥补了“记忆爬虫”法和八邻域法在跟踪分支、断点和“厚”边缘过程中存在的不足。 实验证明该方法效果较好。 关 键 词：区域生长；边界跟踪；爬虫；八邻域 中图分类号：TP 391 文献标识码：A 文 章 编 号：1003-0158(2011)03-0052-05 A New Algorothm for Boundary Tracing SHI Shuang, QU Shi-ru, HE Li ( Department of Automatic Control, Northwestern Polytechnical University, Xi’an Shaanxi 710072, China ) Abstract: For the shortcoming of non-single pixels and broken points in the obtained image boundary, a new algorithm for boundary tracing of dual layer boundary region growing is proposed, to search inner-points and outer-points around center-points, to combine the inner-points with upper outer-points, to conduct continuous tracing with the combined points as the center-points in next search. The algorithm takes into account the one-way search from initial points, and can fill the broken points in one tracing process. Thus, it effectively makes up the defects of memory reptile method and eight neighborhood method in tracing embranchment, broken point and thick boundary. The experiments prove its effectiveness. Key words: region growing; boundary tracing; reptile; eight neighborhood [1] 区域的边缘为图像中灰度变化剧烈的地方， 重建等都要用到边界信息 ，边缘处理的好坏直