基于距离比上下文的形状描述与识别方法

Shape Description and Recognition Approach Based on Distance Ratio Context

基于形状轮廓上的采样点到形状质心的距离,提出了一种距离比上下文形状描述符,用于形状识别和检索。该描述符计算简单,能有效区分不同形状,本质上具有平移、缩放不变性,且在一定程度上能抗部分遮挡和形变。用动态规划算法度量形状比上下文之间的距离,解决了对起始轮廓点的选择问题。在kimia’s-99形状图像数据库中的实验结果表明,该方法在单目标封闭轮廓的形状图像检索中取得了良好的效果。

We suggested a shape descriptor named Distance Ratio Context（DRC） which is based on the distance between sampled points and the centroid of the object.This descriptor has the properties of invariant to scaling and translation essentially and can be calculated easily.It＇s also invariant to deformation and distortion in some ways and can discriminate different shapes effectively.The dynamic programming algorithms were used to measure the distance between DRCs and this mechanism solved the problem of start point choosing on the object contour.The experiments in kimia＇s-99 shape dataset show that this approach,used in image retrieval of shape with a single closed contour,can get favorable results.