摘要
随着多媒体技术的不断发展,每天大量的图像被制造出来,在此情况下,基于内容的图像检索技术发展迅猛。文中提供了一个在基于文本的初始检索结果基础之上,利用伪相关等级信息检索重排序的算法。利用少量的相关性等级计算出其他未标注图像的伪相关性等级,最终根据伪相关等级进行重排序。
With the rapid development of multimedia technology, a large number of digital images swarm into people＇s daily lives. Thus, content based image retrieval （CBIR） is more and more important and image search reranking has become one of the hot techniques in this area. This paper proposes a simple reranking method which utilizes the information of pseudo relevance degree. With only a few relevance degree labels, it calculates the pseudo relevance degrees of other unlabeled images, and then reranks images based on the degrees. The experimental results on the popular MSRA-MM dataset illustrate the effectiveness of the proposed method.
出处
《信息技术》
2014年第10期103-105,109,共4页
Information Technology
关键词
图像检索
相关性等级
重排序技术
image retrieval
relevance degree
visual search reranking
