| Summary: | 碩士 === 義守大學 === 資訊工程學系 === 92 === In recent years, with the fast development of computer technology, the amount of digital images is grown very rapidly. So content-based image retrieval has become an important research issue. We first give an introduction to the state of the art of content-based image analysis and retrieval technology. The purpose of this thesis is to construct an efficient method of natural image retrieval, there fore we focus on color features in our system. The study of relevance feedback is based on the color feature. We propose a mechanism which allows the system to dynamically adjust its distance weighting according to user's feedback. While dimension reduction methods are to guarantee the correctness of feature, relevance feedback makes the retrieval more compatible with human percept matching.
Experimental results indicate the proposed method is very fast and efficient. There fore, the proposed technique demonstrates a satisfactory searching results for the users.
|
|---|
