||A Hybrid Information Ranking System for Web Image Search
Swe Nwe Nwe, Htun ,
Khin Mo Mo, Tunb ,
Yokota, MitsuhiroThi Thi, Zin
232 , 2017-07-31 , 宮崎大学工学部
Nowadays, web image search system is an important part of our daily life due to the tremendous amount of visual information in the World Wide Web (WWW). In order to meet users' satisfaction, many academic researchers has attempted to explore relevant information to the users with high accuracy. However, still a lot of improvements needed to be done. Thus, in this paper, we propose a hybrid information ranking system for web image search in which we use community based platforms with visual features to improve the relevancy between returned images and user intentions. Specifically, we propose a community-specific information ranking algorithm to re-rank the web information by taking user relevance into account. In doing so, we employ a correlated Markov Chain approach along with image similarities and processes of users intentions. Through a series of extensive experiments, we will confirm that the importance of visual factors and community factors, and the effectiveness of the proposed ranking algorithm for Web Image Search.