Research And Implementation Of Content-based Image Retrieval

Content Based Image Retrieval (CBIR) is a technique for retrieving image on the basis of automatically-derived vision feature (color, shape, texture). The task of CBIR is to index, retrieval images automatically and intelligently. On this basis, the user of CBIR system can retrieve images easily, quickly and accurately from image database, and the manager of CBIR system can manage the system without doing a lot of tedious manual work. The image retrieving from database tables and the management of image database has become an urgent research area because of the development of the technology of the database, multimedia and network.This thesis makes extensive and deep research in CBIR technology. The background and key technologies and problem in contented-based image retrieval are introduced. In this paper, we studied the methods of extracting color, shape, texture feature, and similarity measurement. On the basis of this, we proposed a new method which composes of color feature and texture feature. Furthermore, the capabilities of this algorithm is tested and compared between some other classical algorithms. It is possible to use in this content-based image retrieval system. It also makes senses that composes of multi-features get good result than single one.In the end of this paper, a CBIR system for testing retrieval algorithms is developed, which is an experimental frame system. The platform for developing these systems is Windows XP Professional, and the development environment is Visual C++ 6.0. The paper contains two image retrieval experiments. One is based on classical algorithm of content-based image retrieval. The other is an improved algorithm based on both of color and texture features. The results of the experiments are satisfying. It has the theory values and practical meanings to the studying and using in the future of CBIR technology.