一种融合多特征的图像检索方法  

A FUSION METHOD OF MULTI FEATURE IMAGE RETRIEVAL

作　　者：汤琳琳[1,2] 于治楼 

机构地区：[1]山东师范大学信息科学与工程学院,济南250014 [2]山东省分布式计算机软件新技术重点实验室,济南250014 [3]浪潮集团有限公司,济南250101

出　　处：《山东师范大学学报（自然科学版）》2016年第3期32-36,共5页Journal of Shandong Normal University(Natural Science)

摘　　要：基于传统单一特征进行检索的方法已较成熟,本文在提取图像单一特征的基础上,综合三种特征并进行归一化处理进行检索.首先把图像分割成均等的九块并对中间模块加大权值来提取颜色特征,采用改进的不变矩来提取形状特征,最后提取图像的Tamura纹理特征.仿真实验显示,相比于传统单一特征的检索方法,该方法的查全率和查准率更高,更能符合人眼的视觉要求,检索效率更优,因而性能更好.Retrieval method based on the traditional single feature has been more mature, in this paper, based on the extraction of image characteristics of a single feature, method that combined with the three kinds of features and the normalization of processing is used. First, put the image into nine equal pieces and the weights of intermediate modules are increased to extract color feature, then improved invariant moments are employed to get shape character, and Tamura texture features of the image are finally extracted. Simulation experiment shows that compared to the traditional single feature retrieval method, higher recall and precision of the method, better in line with the requirements of the human visual, higher retrieval efficiency, and better performance are obtained.

关 键 词：分块图像 颜色直方图 改进的Hu不变矩 纹理特征 检索 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]