碩士 / 國立成功大學 / 資訊工程學系碩博士班 / 95 / Conventional approaches for content-based image retrieval are always on the basis of the computations of the similarities between user’s query and images via a query-by-example system. Indeed, no matter how the search strategies are powerful, it is very hard to make a precise search within a very short term of query feedbacks. Hence, it motivated us to develop an innovative query refinement technique, namely NQ3, that combines navigation patterns and the hybrid search strategy, namely Q3, with respect to QPM (Query-Point-Movement), QR (Query-Reweighting) and QEX (Query-EXpansion). Generally, the primary difference between our proposed approach and the other contemporary approaches is that we have accomplished excellent image retrieval with considering the problems for visual diversity, exploration convergence, redundant browse. In other words, the major contribution of this paper is that we propose a new Query-Refinement technique to help user make an effective and efficient exploration of images. For efficiency, the expected query formulation scenario is that the intolerant navigations are expected to be prevented by exploiting the navigation patterns hidden in user behaviors. For effect, three query refinement strategies will cooperate to overcome the obstacle about feature diversity. The experimental results reveal that our proposed approach is very effective for query refinement in terms of accuracy through the integration of navigation patterns and Q3.
| Identifer | oai:union.ndltd.org:TW/095NCKU5392047 |
| Date | January 2007 |
| Creators | Wei-Jyun Huang, 黃偉鈞 |
| Contributors | Vincent S. Tseng, 曾新穆 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 70 |
Page generated in 0.0162 seconds