非负矩阵分解耦合视觉多样性的图像检索算法  

Image retrieval algorithm based on non-negative matrix factorization coupled with visual diversity

作　　者：仲宝才[1] 张福泉[2] 徐琳[3] ZHONG Bao-cai;ZHANG Fu-quan;XU Lin(Department of Computer Science and Technology,Chengdu Neusoft University,Chengdu 611844,China;School of Software,Beijing Institute of Technology,Beijing 100081,China;Institute of Innovative Information Industry,Fujian Normal University,Cangshan 350007,China)

机构地区：[1]成都东软学院计算机科学与技术系,四川成都611844 [2]北京理工大学软件学院,北京100081 [3]福建师范大学创新信息产业研究所,福建仓山350007

出　　处：《计算机工程与设计》2018年第10期3201-3207,共7页Computer Engineering and Design

基　　金：国家教育部博士点基金项目(20121101110037);四川省教育厅课题基金项目(15ZB0359)

摘　　要：为解决图像检索中视觉特征与语义概念间的语义鸿沟现象,导致其检索不准确的问题,设计非负矩阵分解耦合视觉多样性的图像检索算法。对图像进行非负矩阵分解(non-negative matrix factorization,NMF),得到图像的基矩阵与系数矩阵;在NMF过程中嵌入整体视觉多样性,利用图像之间的差异描述视觉多样性;引入两个正则项对其进一步优化,根据分解因子的乘积完成缺失标签的重构,获取标签图像矩阵;利用图像矩阵的标签语义距离进行相似性度量,完成图像检索。实验结果表明:在Flickr25K、VOC2012数据集中,与当前检索技术比较,所提算法具有更高的检索精度。To solve defect of low retrieval accuracy induced by semantic gap between visual features and semantic concepts in image retrieval,an image retrieval algorithm based on non-negative matrix factorization coupled with visual diversity was designed.The basis matrix and coefficient matrix of the image were obtained using non-negative matrix factorization to process the image.The visual diversity was described using the difference of image whole visual diversity was embedded in the process.The deletion tag was reconstructed according to the product of the factorization factor for obtaining tag image matrix by introducing two regularization terms.The similarity measure was carried out by calculating the tag semantic distance of the image matrix to complete the image retrieval.Experimental results show that the proposed algorithm has higher retrieval accuracy in Flickr25K,VOC2012 data sets compared with the current retrieval techniques.

关 键 词：图像检索 非负矩阵分解 视觉多样性 标签图像矩阵 正则项优化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]