In recent years, with the comprehensive popularization of computer network, multimedia information retrieval has developed rapidly. The Digital Library scheme of IBM converts physical information into digital multimedia form and sends it safely to users all over the world through the network. Natural language query and concept query filter the information returned to users, making the positioning of relevant data simpler and more accurate. The aggregation function organizes the query results together, so that users can simply identify and select relevant information. The summary function can summarize the main points of the query results, so that users can determine the information they want to find without viewing all the text. IBM Almaden Research Center launched QBIC system. This system opens up a new field of image information query. Images can be queried by color, grayscale, texture and position. The query requirements will be expressed graphically, such as selecting colors from the color table, or selecting the texture of the image from the example image. The query results can guide the sub sequence query according to the relevant sequence. This method enables users to filter and determine visual information more quickly and easily. With the support of the "863" high-tech project, we study the feature-based MIRES(Multimedia Information REtrieval System). 

Operating Environment：

The client and web server of the product are developed in Java language, the application server is developed in c++ language, and then the middleware server CORBA is used for integration. The application server can run on Windows2000, Windows NT and windows server. The web server can run on various mainstream operating systems, including windows, UNIX, Linux, Macintosh, etc. It is required to have a Java running environment (Java virtual machine JVM), such as jdk1.3 or above. In addition, it must have a middleware CORBA server, such as VisiBroker, Orbix, etc. The supporting software of web server can support any platform of jsp/servlet, such as Tomcat. The client only requires a web browser that supports Java, such as ie4.0. Hardware requires PII, 256M memory or above, or mainstream models such as SUN or HP workstations.

Major Functions：

On the MIRES client, users can complete the following functions:

(1) Provide the sample image to be retrieved and submit it to the image server;

(2) Provide the semantic keywords of the image to be retrieved and submit them to the image server;

(3) Users can interactively mix the above two modes to retrieve the desired image.

(4) For the sample image retrieval method, the user is allowed to set the importance coefficient of various features;

(5) For the returned result image group, the user can make the retrieval more in line with the user's requirements through interactive feedback.

On the server side, receive the retrieval request returned by the client, analyze it, select the query strategy, and start the retrieval engine for retrieval; Sort the query results and return them to the client. The server also has the functions of managing the retrieval process and image database. Interface and encapsulation function, which can well interface and encapsulate various existing software or programs, including plug-in, embedded, dynamic connection library and other interfaces. The operation platform can provide network operation mode, that is, a platform can be distributed on multiple computers, and multiple platforms can also communicate with each other and access services. There should be friendly and efficient communication mode between client and server, and different protocols can be selected for communication.

Key Technology：

● Feature based image information retrieval: image information has rich connotation and gives people intuitive, vivid and vivid feelings. It is an important information form in multimedia information. This paper focuses on the retrieval of image information according to the characteristics of color, shape, texture and so on.

● Chinese information full-text retrieval: Research Chinese information full-text retrieval methods, with Boolean retrieval, truncation fuzzy matching retrieval, complete string matching retrieval, location adjacent retrieval and other retrieval mechanisms. Use the relevant index mechanism to improve the retrieval efficiency.

● Creation and acquisition of content: convert the data in physical form into digital information, and compress and convert it. Due to the use of a large number of images and videos, the data volume of multimedia information carrier is hundreds of times larger than that of traditional text-based single media. Data compression and restoration has become a key technology in the development of multimedia.

● Information storage management: the storage objects can be text, graphics, image digital information, etc. Study the methods of large-scale information organization and management, and explore the multidimensional spatial index method to improve the effectiveness of information management and retrieval efficiency. The information storage management adopts the client / server mode.

● Cross platform client-side retrieval: cross platform client-side retrieval technology is to solve the problem of unified graphical retrieval interface on various software and hardware platforms under the network environment (especially the internet/intranet environment). In the network environment, the software and hardware configurations of different retrieval clients are very different. The cross platform client retrieval technology will make these differences transparent to users. Users on different platforms will use a unified graphical retrieval interface for retrieval work.