一种基于区域分割的图像检索方法  

A method of image retrieval based on region segment

作　　者：郑声恩[1] 岑炜桦[1] 叶少珍[1] 

机构地区：[1]福州大学数学与计算机科学学院,福建福州350002

出　　处：《福州大学学报（自然科学版）》2006年第6期821-825,共5页Journal of Fuzhou University(Natural Science Edition)

基　　金：863计划资助项目(2002AA111090)

摘　　要：提出了一种基于区域分割的图像检索方法:首先用改进的基于一致性分析的空间分割算法对图像进行区域分割,然后提取各区域的视觉特征,再对查询图像和目标图像中各区域的视觉特征进行多对多的匹配,采用最小平均准则和高斯归一化加权求和得到图像间的相似度.通过与基于全局特征的检索结果比较表明,该方法可以产生比较满意的结果.The method of region based image retrieval is proposed in this paper. Firstly an improved method based on local homogeneity analysis is used to segment the image, and the vision features are extracted. Secondly the features between the query image and the target images are matched on many to many, so the distance between the images is obtained according to the minimal weighted mean criterion, by which comparability measurement is analyzed. The experiment shows that the satisfactory result can be got by the method above.

关 键 词：图像检索 图像分割 基于区域 相似度 

分 类 号：TN918[电子电信—通信与信息系统]