| Summary: | 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 94 === In recent years, a number of research works have been proposed for content-based image retrieval with different features. Theses works usually use the content of images, low-level features to calculate the similarity between images that hope to find the similar images and retrieve image database efficiently. Although some novel content-based image retrieval systems may provide good retrieval results, the low-level features cannot entirely represent the high-level human sense in retrieval. Hence, the retrieval process is not effective. In fact, the concepts hidden in the images are often the key targets for image retrieval from the viewpoint of human sense. In this paper, we propose a new approach named Intelligent Concept-Oriented Search (ICOS) that can catch the high-level concepts of the users for performing content-based image retrieval by using a data mining approach. Through experimental evaluation, the proposed approach can provide fast and accurate image retrieval in user-friendly interface. We also develop new skills like the methods of intelligent concept-oriented search for content-based image retrieval, object annotation, association rules between objects, and the ranking methods for ontology concept.
|
|---|
