| Summary: | 碩士 === 淡江大學 === 資訊工程學系 === 88 === Content-based image retrieval has become more desirable for developing large image database. We propose a new method of retrieving images from an image database in this research plan. We combine the color, shape and spatial relation features of a picture to index and measure the similarity of images. For any color-based image retrieval system, the key issues are the selection of color space and reducing the resolutions of the color histograms in order to decrease the computing complexity and to increase the performance of similarity measurement. In this plan, several color spaces that widely used in computer graphic were discussed and compared for color clustering. In addition, we propose a new automatic indexing scheme of image database according to our clustering method, which could filter the image efficiently. As a technical contribution, a Seed-Filling like algorithm that could extract the shape and spatial relationship feature of image is proposed. Also, we propose a shape normalization algorithm to increase the precision of image retrieval. And, we extend temporal interval relation by means of a complete analysis for spatial computation of the image. Besides, the system is incorporated with a friendly graphic user interface, which allows the user to retrieval the image easily.
|
|---|
