The ever increasing demand of large scale multimedia applications has made the development of effective multimedia database systems an important direction of research. Distributed architecture is widely used in this setting due to the large volume of data involved. This introduces a new challenge to adaptive multimedia data retrieval. Several issues have to be considered such as how trust can be propagated in an autonomous environment, and how to aggregate the results from individual databases and work out an overall ranking. In this paper, we propose a multimedia data retrieval method in a heterogeneous distributed database system. An agent reputation model that is based on social network analysis methods is employed for handling trust in the heterogeneous environment. An adaptive data retrieval method is utilized for an automated querying process based on the reputation model. Sub-group analysis is conducted for better support of collaborative ranking and community-based search. Experiment results show good performance of the proposed method.