| Content Based Image Retrieval (CBIR) has been a very active research area since 1990's, with the thrust from two major research communities, Database Management and Computer Vision. The traditional approach of Image Retrieval is based on the technology of Database Management Systems (DBMS), with the cost of heavy burden of manual annotation. The proposed Content Based Image Retrieval, however, is a new approach based on Computer Vision, Pattern Recognition which perform the computer-centered image retrieval according to the content of images.From the point of view of establishing a practical CBIR system, this paper analysis and introduced all the methods used in our system. They are:First, the basic techniques for establishing the pratical CBIR system, such as, image processing, the extracting of visual features from image and the computing of similarity between visual features of different images and furthermore the computing of distance between images.Second, the advanced approach to improve system performance. They can be divided into two categories: one for improving the accuracy of retrieval and the other for improving the speed of retrieval, in our system, former method is relevance feedback and the latter is vector quantitation.Later on, using semantic feature of images in CBIR system. The model for combining semantic feature and visual features of images into one system is proposed.Finally, other techniques that is necessary to our system, such as Oracle Database Design, User Interface and techniques for design a web-based image retrieval system.In conclusion, the purpose of this paper is trying to be a comprehensive article for CBIR system: Besides concentrate on our system, we introduced some other famous systems too. In addition to presented some important methods used in our system, we also get to the bottom of their theoretical origination and background knowledge. There are also some detailed analyses of typical algorithms and necessary experiment results in this paper. |
PDF Full Text Request