<XML><RECORDS><RECORD><REFERENCE_TYPE>3</REFERENCE_TYPE><REFNUM>9314</REFNUM><AUTHORS><AUTHOR>Feng,Y.</AUTHOR><AUTHOR>Jose,J.</AUTHOR></AUTHORS><YEAR>2008</YEAR><TITLE>A Hybrid Approach for Classification Based Multimedia Retrieval</TITLE><PLACE_PUBLISHED>3rd international conference on Semantic and Digital Media Technologies, Koblenz, Germany</PLACE_PUBLISHED><PUBLISHER>N/A</PUBLISHER><LABEL>Feng:2008:9314</LABEL><ABSTRACT>In this paper, we investigate an approach combining classification and retrieval for the Content Based Image Retrieval (CBIR). Traditional CBIR mainly relies on local features in an image. However, this approach is inadequate effective for semantic search. Here, we focus on bridging the gap between low level image features and high level features by applying a coarse to fine strategy of query learning and retrieval. The proposed algorithm firstly classifies the data set into a set of predefined categories using image global features. This is followed by a retrieval process to extract the most similar images for the query using combined image local features. This system features in (i) saving the computation cost by applying and (ii) improve the retrieval effectiveness. The experiments are made using different image collections, and the results show that the processing time and the precision rate is improved comparing to existing methods.</ABSTRACT></RECORD></RECORDS></XML>