期刊名称：International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN：2320-9798
电子版ISSN：2320-9801
出版年度：2014
卷号：2
期号：2
出版社：S&S Publications
摘要：Searching a image from search engines and social networks are providing results based on thecomments, clarification and some other information about the images. The results will be generated easily, Due toineffective similarity mining the systems may suffer from the accuracy. This paper tells how to overcome the problemof image mining based on effective meta informations and semantic similarity measures. The semantic similaritycontains both textual and visual similarity measures. To resolve the mentioned problems, Content-based image retrieval(CBIR) avoids the use of textual descriptions and instead retrieves images based on the content similarities like colours,shapes, textures etc to the user-specified image features. We proposed a method named Content-Based RelevanceResponse (CBRR) is used to achieve the high retrieval quality by using the discovered patterns. In case of efficiency,the patterns are mined from the user query log can be viewed as the shortest paths to the user space. According to thepatterns, the users can obtain a set of relevant images in an online query refinement process. Similarity for the searchcan be done by using meta tags, shape/region attributes, and colour distribution in images
