Multimedia Retrieval System Based On Content

In the past years, there is a quick development of computer application in a great deal of vocations, and many data-bases are built in most domains, the quantity of data increased every day, and some of them are issued on the Web; all kinds of multimedia data of Web exploded, and the Web is growing up to a data warehouse including of text, images, audio type of data. In the face of a so huge warehouse, data mining is absolutely necessary, the retrieval systems for multimedia data-bases are more necessary .Moreover, most of retrieval systems are still based on description now, they set up index and identification for image data and retrieval also bases on image description (key words, title, etc.), this kind of index is random, and the quality of the retrieval result is not every well. So this paper recommends a retrieval system, which based on content. This system extracts the image character itself (color, texture, etc.) to set up index and identification, computes the similarity of the characters when retrieval and makes use of the query refinement method,the quality of the query result is all right.Above all, this paper introduce the theory of data mining, especially about Web data mining and multimedia data-base; and in face of the retrieval system which based on content, recommend the concept of multimedia object model, query tree, query refinement, etc. Test the precision and recall of this system, retrieval the documents of different length and multimedia type, and get a perfect result. Finally, certificate the necessary of the query refinement.