The query conditions are imprecise or uncertainty in multimedia information retrieval. So it is impossible to get completely information by using traditional retrieval method and new techniques must be developed in order to get the best similar retrieval results under the inaccurate or incomplete retrieval conditions given. In this paper, a fuzzy semantic distance model and a concept of fuzzy similarity are first proposed and fuzzy query processing in fuzzy space is described; then, the algorithm of information retrieval in multimedia databases based on fuzzy similarity is provided; finally, an example is used for explanation the algorithm.