| Summary: | 碩士 === 東吳大學 === 資訊科學系 === 90 === With the fast development of internet and computer technology in recent years, the use of visual information on the WWW has become more popular. There has been an increase in use of images and the associated image databases are getting larger. The problem of conducting the image retrieval has attracted plenty of the attention from researchers of information technology. Traditional information retrieval techniques, such as text-based information retrieval, have provided access to unstructured textual documents. Retrieving of images is more complicated. It involves extracting the most important or remarkable characteristics of images, these characteristics are organized in some fashion in order to access them efficiently. Extracting images of their content is really an arduous job. Text information is always structured at the syntactic level, we can use machine to supporting automatic extraction of keywords for content based indexing of text. But image structuring is partial and subjective, therefore it cannot be recovered at a syntactic level.
While there are limitations in current techniques for subject access to image data, CBIR (content-based image retrieval) nowadays has gotten more attraction. CBIR is the name of technique that has been widely used to describe the process of retrieving desired images from a large collection on the basis of features (such as color, texture and shape) that can be automatically extracted from the images themselves. In this paper, a novel idea in designing a CBIR system was proposed, which makes use of MOT (Mesh Of Trees) based search strategy and dual image signatures to provide better query result and efficiency requirements. Experimental results show that the proposed architecture demonstrates its scalability and performance in these aspects.
|
|---|
