标题:Automatic image annotation by visual topic discovering and web mining
作者:Liu, Zheng ;Ma, Jun
作者机构:[Liu, Zheng ;Ma, Jun ] School of Computer Science and Technology, Shandong University, Ji'nan 250101, China;[Liu, Zheng ] School of Computer Science a 更多
通讯作者:Liu, Z
来源:Journal of Information and Computational Science
出版年:2010
卷:7
期:2
页码:365-371
关键词:Automatic image annotation; Hash code; Image similarity; LDA; Visual topic
摘要:A novel automatic image annotation approach based on visual topic discovering and web mining is proposed in this paper. Firstly, visual topics are learned from corel5k dataset using Latent Dirichlet Allocation (LDA) model. Given an unlabeled image, we apply LDA to discover visual topic of this image, and then generate initial annotations from the visual topic which is most related to the unlabeled image. Secondly, initial annotations are submitted as query keywords to image search engine, and then visual similar images are found according to hamming distance from the searching results by mapping the high-dimensional image visual features into hash codes. Next, extended annotations are mined from the text descriptions (titles, URLs and surrounding texts) of these images. Finally, combining initial annotations and extended annotations, final annotations are obtained. Experimental results show the effectiveness and efficiency of the proposed approach. Copyright © 2010 Binary Information Press.
收录类别:EI;SCOPUS
资源类型:期刊论文
原文链接:https://www.scopus.com/inward/record.uri?eid=2-s2.0-77952525303&partnerID=40&md5=235d667b533f67b751893155fb85a6c2
TOP