基于形状上下文的形状匹配  

Shape Matching by Shape Context

作　　者：申家振[1] 张艳宁[1] 刘涛[1] 

机构地区：[1]西北工业大学计算机学院,陕西西安710072

出　　处：《微电子学与计算机》2005年第4期144-146,共3页Microelectronics & Computer

摘　　要：提出了用于形状匹配的自适应边界点选取的形状上下文方法。首先,根据目标形状,自适应地确定边界点个数;然后,利用选取的边界点,按照本文方法计算该目标的形状直方图;最后,利用形状直方图计算不同目标之间的相似度。基于手写体数字的试验表明,形状上下文在有效描述目标的形状特征的基础上,准确地计算了不同目标之间的相似度,达到了满意的匹配结果。A method for shape matching called shape context by choosing the edge points self-adaptively is proposed in this paper. Firstly,self-adaptively choose the edge point according to the object,and then,calculate the shape histogram of one single object based on the edge points using this method introduced in this paper,finally,compute the similarity between different objects using shape histogram?Experiments on handwritten digits show that this method computes the similarity between different objects precisely and gains desirable matching results on the basis of efficient description of the shape.

关 键 词：形状匹配 形状上下文 手写体数字识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]