基于形状特征的图像搜索研究

A Study of Shape-Based Image Retrieval

基于内容的图像搜索可以分为特征提取,对象表示和对象匹配。在特征提取中有色彩,纹理,形状和空间关系等特征,而形状特征能给人们带来非常直观的信息。在形状特征的边缘提取中采用Prompt edge detection方法来获取图像边界,并与Sobel算子边缘提取法进行比较。在对象表示中采用边缘点的爬山序列,爬山序列能在对象移动,旋转和缩放后保持一致性,是一种较好的对象特征表示方法。

Content-based image retrieval work includes feature selection,object representation,and matching.There are color,texture,shape and spatial relationship.Among these features,the shape can bring people attractive visual information.We take the method of Prompt Edge Detection（PED）as edge point detection,which is compared with the Sobel edge detection method.In the processing of object representation,we use the mountain-climbing sequence（MSC）,which is invariant to translation,rotation,and scale.The MSC is a good method for object presentation.