Abstract: With rapid growth of network image data, the effect of traditional image retrieval can; t meet current needs. To solve this problem, we propose an image retrieval method based on iterative quantization (ITQ)and deep belief network (DBN). The method of image retrieval firstly constructs a model of deep confidence network, which is formed by stacking three layers of restricted Boltzmann machines. Then, we adopt the constructed deep confidence network model to extract image features by using iterative quantization in the Greek method, and the extracted features are binary coded. Finally, the model is verified by MNIST, CIFAR-10 and Corel-1000 datasets. The results show that the proposed method has better retrieval performance than the current mainstream methods. In addition, the method has achieved good results in the retrieval of breast dataset DDSM and lung nodule CT image dataset LIDC-IDRI.