In this paper, we proposed a content-based multimedia analysis/retrieval system basing mainly on the MPEG-7 defined features. Some new and specific features are also included for supporting various requirements of different applications. The proposed system is capable of extracting a variety of features from different kinds of media data, such as videos, audios, and images. The system design is well modularized, so that the tasks of system development and feature construction can be done independently and separately. In other words, adding a new feature into the system can be done without modifying the structure of the system. In addition, a user can also input some hints about what he or she is looking for into the system, and the system will search through the database and return the best matched candidates. On the basis of this well-modularized system, a lot of interesting applications, such as scene change detector, video news program analyzer, Logo verification system, fish shape...