The CBIR Research Based Four Features

The image is defined as the representing form and storing carrier of information in visual. With the development of multimedia, computer, the telecommunication technique, and internet, the source of the images is more and more extensive, the image database has been became the main means to organize, express, store, search, query, and employ these a great deal of information. The image retrieval is a key problem at the image database research. At recent, the content-based image retrieval (CBIR) has become one of key issues that studied at home and abroad.This paper adopts four methods to search images, and uses dynamic weight methods to rank the searched images. In statistic field, we adopt principal component analyses method and singular value decomposition method. In image color distribution field, we offer multiple resolving power dynamic illumination main color algorithm. In image content shape figured by corner points, we offer a new algorithm which uses optimization pair set thinking in graph and improved Hausdorff algorithm to measure the similarity of two point sets, and propose a dynamic weight adjust algorithm based information entropy. Multiple features retrieve method and dynamic rank method improved the stabilization of retrieve result.