Optimal estimation of descriptor scales for multimedia retrieval

Doulamis, N

dc.contributor.author	Doulamis, N	en
dc.date.accessioned	2014-03-01T02:44:53Z
dc.date.available	2014-03-01T02:44:53Z
dc.date.issued	2007	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/31997
dc.subject	Information Need	en
dc.subject	Multimedia Retrieval	en
dc.subject	Optimal Estimation	en
dc.subject	Relevance Feedback	en
dc.subject.other	Control theory	en
dc.subject.other	Feedback	en
dc.subject.other	Image analysis	en
dc.subject.other	Information retrieval	en
dc.subject.other	Military data processing	en
dc.subject.other	Neural networks	en
dc.subject.other	Best fit	en
dc.subject.other	descriptor	en
dc.subject.other	Descriptors	en
dc.subject.other	Dynamic modifications	en
dc.subject.other	Information needs	en
dc.subject.other	International (CO)	en
dc.subject.other	Multimedia contents	en
dc.subject.other	Multimedia interactive services	en
dc.subject.other	Multimedia retrieval	en
dc.subject.other	Optimal estimations	en
dc.subject.other	Relevance feedback (RRF)	en
dc.subject.other	Retrieval (MIR)	en
dc.subject.other	Retrieval performance	en
dc.subject.other	retrieval processes	en
dc.subject.other	User's preferences	en
dc.subject.other	Multimedia services	en
dc.title	Optimal estimation of descriptor scales for multimedia retrieval	en
heal.type	conferenceItem	en
heal.identifier.primary	10.1109/WIAMIS.2007.64	en
heal.identifier.secondary	http://dx.doi.org/10.1109/WIAMIS.2007.64	en
heal.identifier.secondary	4279186	en
heal.publicationDate	2007	en
heal.abstract	Relevance feedback schemes use a given set of descriptors and at each feedback iteration they just modify the importance of each descriptor to the retrieval performance. This implies that the same type of descriptors is used for all feedback iterations. Therefore, such approaches cannot improve the retrieval performance beyond the efficiency of the descriptors used to represent the multimedia content. In this paper, an alternative concept is to adopt a progressive descriptor organization and to allow dynamic modification of the type of descriptors (expansion or shrink) at each retrieval iteration by exploiting the current user's information needs. In other words, descriptors that best fit the current user's preferences are expanded into more details, in constant to the descriptors that are far way from the user's needs which are shrinking. In this case, instead of using a large set of descriptors, most of them inappropriate with respect to the current user's information needs, a small number of dominant descriptors is selected for the retrieval process. Thus, the proposed approach significantly increases the precision performance by discarding descriptors that yield noise in the retrieval. © 2007 IEEE.	en
heal.journalName	8th International Workshop on Image Analysis for Multimedia Interactive Services, WIAMIS 2007	en
dc.identifier.doi	10.1109/WIAMIS.2007.64	en