期刊名称：International Journal of Computer Science and Information Technologies
电子版ISSN：0975-9646
出版年度：2014
卷号：5
期号：4
页码：5935-5938
出版社：TechScience Publications
摘要：Image re-ranking, as an efficient way to advance the results of web-based image search, has been implemented by present commercial search engines such as Bing and Google. For a given query keyword, a group of images is first retrieved based on textual information. From the group of images, the user selects the query image and the rest of images are re-ranked based on their visual similarities with the query image. A main dispute is that the similarities of visual features do not well correlate with semantic meanings of images which understand users’ search purpose. This dispute is addressed by Keyword Query expansion an effective way to improve the performance of information retrieval systems by adding additional relevant keyword terms to the original queries. This paper presents a new method for keyword query expansion based on query logs mining. This method first calculates the ambiguity degree of the query by developing the user logs. The result measures the quality of the keyword query and decides the expanded length of the query. Next, the information of clicked images is used to create the correlations of the keyword queries, and the high-quality expanded keyword terms are selected from the past queries with the investigation of relation between queries and Images. This approach provides an effective way to avoid the problem of query drift by reducing the irrelevant expansion terms. The eexpanded keywords are used to widen the image pool to hold more relevant images. To further improve content-based image reranking, expanded keywords are used to enlarge the query image to multiple positive visual examples from which new query specific visual and textual similarity metrics are learned. Experimental result of the present work achieves better result when compare with the existing work
