基于轮廓分割的形状匹配新方法  

A new method of shape matching by contour segmentation

作　　者：王爱平[1] 余江[1] 江丽[1] 张芹芳[1] 

机构地区：[1]安徽大学计算机科学与技术学院,安徽合肥230601

出　　处：《微型机与应用》2013年第8期88-90,共3页Microcomputer & Its Applications

摘　　要：提出了基于轮廓分割的形状描述方法,用来进行形状匹配。首先,使用离散曲线演化将形状轮廓简化成具有N个顶点的多边形,并通过去除其中的凹点得到多边形凸点的集合;然后,对由凸点形成的轮廓片段提取角度和曲率特征,得到形状描述子;最后通过匈牙利算法进行轮廓片段之间的距离矩阵的最小代价和计算,从而得到两个形状之间的相似度。基于MPEG-7图像库上的实验证明了本文方法能很好地反映形状的特征,具有较好的匹配效果。A shape deseription method based on contour seglnentation, used for shape matching, is proposed in this paper. Firstly, simplify the contour of the image based on discrete curiae evolution into a polygon with N vertices, and remove pits of the vertices to obtain a collection of polygonal bumps, and then, extraction angle and curvature characteristics of the contour fragments between the bump to get shape descriptors, finally, by the Hungarian algorithm for caleulation of the minimum eost of distanee ma- trix between the contour fragments, and obtain the similarity of the two shapes. Experiments on MPEG-7 image library＂ prove that this method can express shape feature and gain desirable matching results.

关 键 词：形状匹配 轮廓分割 离散曲线演化 形状的相似性度量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]