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：5
页码：6367-6372
出版社：TechScience Publications
摘要：Search re-ranking is considered a typical thanks to boost retrieval exactitude. The matter yet isn't trivial particularly once there square measure multiple options or modalities to be thought-about for search, which frequently happens in image and video retrieval. This paper proposes a replacement re-ranking algorithmic program, named data processing based mostly on re-ranking, that reinforces the mutual exchange of knowledge across multiple modalities for up search performance, and following the philosophy that sturdy activity modality might learn from weaker ones, whereas weak modality will have the benefit of interacting with stronger ones. The prevailing strategies for image search re ranking suffer from the undependableness of the assumptions below that the initial text-based pictures search result. The ensuing pictures contain additional tangential pictures. Thus the re ranking idea arises to re rank the retrieved pictures supported the text encompassing the image and information and visual feature. Variety of strategies square measure compared for this re-ranking. The top-ranked pictures square measure used as (noisy) coaching information and SVM visual classifier is learned to boost the ranking additional. We have a tendency to investigate the sensitivity of the cross-validation procedure to the present clangorous coaching information. The principal novelty of the methodology is in combining text/metadata and visual options so as to realize a very automatic ranking of the pictures. General Terms-Image retrieval and re-ranking.
关键词：Image retrieval; image search re-ranking; data;mining; visual information retrieval.
