{"title":"Iterative bilinear similarity measure learning for CBIR via a minimization of a local sensitivty","authors":"Shao-Hua Yin, Jincheng Li","doi":"10.1109/ICMLC.2012.6358943","DOIUrl":null,"url":null,"abstract":"With the fast development of photographic device and the Internet, millions of images have been uploaded to the Internet. So, there is an increasing needs of image retrieval method for large scale databases. Content Based Image retrieval (CBIR) is very useful when user wants to find some images that are similar to a given image in both visual and content. In this paper, we first summarize the research development of this field. Secondly, OASIS [5] (Online Algorithm for Scalable Image Similarity) is introduced. However, OASIS focuses on the difference of similarity values between relevant and irrelevant image pairs while ignores the similarity value between highly similar images. So we proposed an improvement to OASIS via a minimization of local Q-neighborhood's sensitivity. It provides a better generalization and retrieves more near duplicate and highly similar images. The proposed improvement is compared with the original OASIS method and yields a better performance.","PeriodicalId":128006,"journal":{"name":"2012 International Conference on Machine Learning and Cybernetics","volume":"95 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2012-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2012 International Conference on Machine Learning and Cybernetics","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICMLC.2012.6358943","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
With the fast development of photographic device and the Internet, millions of images have been uploaded to the Internet. So, there is an increasing needs of image retrieval method for large scale databases. Content Based Image retrieval (CBIR) is very useful when user wants to find some images that are similar to a given image in both visual and content. In this paper, we first summarize the research development of this field. Secondly, OASIS [5] (Online Algorithm for Scalable Image Similarity) is introduced. However, OASIS focuses on the difference of similarity values between relevant and irrelevant image pairs while ignores the similarity value between highly similar images. So we proposed an improvement to OASIS via a minimization of local Q-neighborhood's sensitivity. It provides a better generalization and retrieves more near duplicate and highly similar images. The proposed improvement is compared with the original OASIS method and yields a better performance.