摘要
为提高骨架提取算法的适用性,提出一种新型的骨架提取算法.通过对对象的边界元素按照空间距离顺序标号,求出对象内部像素的边界差,由边界差得到8连通的骨架分层.为提高算法的处理速度,提出前向分层和反向跟踪两个过程的骨架细化方法,用向量差Vd和长宽比(LWR)两个参数及支持向量机(SVM)分类器对冗余的骨架分支进行剪枝处理.试验结果表明,该算法提取的骨架具有很好的连通性,尤其适用于提取对象狭长区域的骨架线.
A novel algorithm for skeleton extraction is proposed. By numbering object＇s border elements on spatial position, the border gap of inner pixel of the object is calculated, and an 8-connected medial- axis hierarchy is derived by the border gap, also a thinning method including slicing and counting is proposed to improve the processing speed. Branches with minor importance are truncated by vector divergence （Vd）and length-width ratio （LWR） with support vector machine （SVM） classifier. Experiments demonstrate that the derived skeletons have satisfying connectivity, especially in long and narrow area.
出处
《东华大学学报（自然科学版）》
CAS
CSCD
北大核心
2013年第2期207-213,共7页
Journal of Donghua University(Natural Science)
关键词
骨架
中轴
边界向量
边界差
支持向量机(SVM)
内积
距离变换
skeleton
medial-axis
border vector
border gap
support vector machine （SVM）
inner product
distance transform
