In this paper, a new materials image retrieval methodology based on dynamic Bayesian network (DBN) is proposed to overcome certain drawbacks inherited in previously proposed CBIR methods. Firstly, the DBN structure and initial parameters is constructed according to the prior knowledge. Secondly, a training phase is conducted for updating DBN parameters through using the feedback information. The training cycles can be stopped until the retrieval accuracy is adequately high. Experimental results on 10,000 images demonstrate the effectiveness of the proposed methodology.