Integrating textual and visual information for cross-language image retrieval
Journal
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Journal Volume
3689 LNCS
Pages
454-466
Date Issued
2005
Author(s)
Abstract
This paper explores the integration of textual and visual information for cross-language image retrieval. An approach which automatically transforms textual queries into visual representations is proposed. The relationships between text and images are mined. We employ the mined relationships to construct visual queries from textual ones. The retrieval results of textual and visual queries are combined. We conduct English monolingual and Chinese-English cross-language retrieval experiments to evaluate the proposed approach. The selection of suitable textual query terms to construct visual queries is the major concern. Experimental results show that the proposed approach improves retrieval performance, and nouns are appropriate to generate visual queries. © Springer-Verlag Berlin Heidelberg 2005.
SDGs
Type
conference paper