There is no doubt that content-based information retrieval technology is an emerging trend for digital library development and will be an important complement to the traditional text-based retrieval technology. The ideal CBIR system can semantically understand the information in a digital library, and render users the most desirable data. However, the machine understanding of semantic information still remains to be a great difficulty. Therefore, most current research projects, including those discussed in this paper, deal with the understanding and retrieval of lower-level features or physical features of multimedia content. Certainly, as related disciplines such as computer vision and artificial intelligence keep developing, more researches will be done on higher-level feature-based retrieval.