分层多特征融合图像检索方法研究  

Research on stratified multi features fusion based on image retrieval method

作　　者：潘红艳[1] Pan Hongyan(Zhejiang Business Technology Institute,Ningbo 315012,China)

机构地区：[1]浙江工商职业技术学院,浙江宁波315012

出　　处：《无线互联科技》2018年第8期68-70,73,共4页Wireless Internet Technology

基　　金：浙江省教育厅科研项目;项目编号:Y201738641

摘　　要：为了克服单一图像特征匹配的不足,文章提出了分层多特征融合的图像检索技术。提取图像特征时,先将RGB图像映射到HSV空间,再分别提取待检索图像和数据库图像的颜色特征和纹理特征。提取颜色特征时,采用颜色矩特征,提取图像的纹理特征时采用灰度共生矩阵的方法。在颜色特征层和纹理特征层,分层分别计算检索图像和数据库图像的相似度,对分层计算出的相似度采用一定的融合策略得到最终的相似度,依据相似度大小对检索结果进行排序和输出。在自建数据库上的实验表明,该方法对提高图像的检索性能有一定的效果。To overcome the shortage of single image feature matching,this paper proposes a stratified multi feature fusion image retrieval technology.When the image features are extracted,the RGB image is first mapped to the HSV space,and then the color and texture features of the images to be retrieved and the database image are extracted.When the color feature is extracted,the color moment feature is used to extract the texture features of the image,and the gray level co-occurrence matrix is used.In the color feature layer and texture feature layer,the similarity between the retrieved image and the database image is calculated by stratification,and the final similarity is obtained by a certain fusion strategy based on the similarity calculated by the layering.The retrieval results are sorted and output according to the similarity degree.The experiment on the self built database shows that this method has a certain effect on improving the retrieval performance of the image.

关 键 词：图像映射 颜色矩 灰度共生矩阵 分层融合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]