A new strategy for skeleton pruning

Pattern Recognition Letters(2016)

引用 15|浏览4
暂无评分
摘要
Pruning does not require to be iterated, since it removes in one step the longest possible concatenations of branches.Information on branch significance is collected along the skeleton before performing pruning.Pruning thresholds are suggested as a function of the object's size. A new pruning algorithm is introduced to simplify the structure of the skeleton of 2D objects, without affecting significantly the representative power of the skeleton. The concatenations of skeleton branches originating from the end points of the skeleton are examined while building a hierarchical skeleton structure. Skeleton branches that can be interpreted as peripheral branches at any hierarchical level are concatenated with the adjacent skeleton branches that were interpreted as peripheral branches at the immediately previous level. A concatenation extends in the hierarchy for a number of levels related to the number of branch points connecting the successive branches along the concatenation itself. The most internal branch point up to which a concatenation can be pruned without affecting the representative power of the skeleton is determined by using significance measures able to evaluate the loss in object recovery produced by pruning that part of the concatenation. Pruning is performed by removing all the so identified longest prunable concatenations, provided that topology is maintained Display Omitted
更多
查看译文
关键词
2D Skeleton,Skeleton branches,Hierarchical structure,Pruning
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要