| Summary: | 碩士 === 國立中興大學 === 資訊管理學系所 === 104 === Traditional text-based image annotation and retrieval approaches combine the visual features and text to deal with the semantic gap problem in content-based image retrieval. However, the terms or labels annotated on images might be ambiguous. Without being clearly defined, retrieval system is difficult to understand the real intention of users’ requests. To address this problem, the state-of-the-art in image annotation and retrieval research starts to incorporate the ontology and technique of Semantic Web. Most of existing ontology-based image annotation methods has some drawbacks. First, they are inconvenient and time consuming because some approaches need to annotate image manually. Second, they require well domain ontology but it is difficult to construct. Third, to fit all the requests for user query, users add annotations on images as their wish, but the amount of the annotations is not as a consider factor. As our observation, many annotations are not necessary to image retrieval. To address the problems of image retrieval, we propose two schemes based on ontology and linked open data, automatic semantic image annotation model and image retrieval model with dynamic semantic query expansion approach. It can increase the precision of image retrieval to fit the requests of users’ queries.
|
|---|
