一种深度置信网络和迭代量化的图像检索方法

Deep Belief Network and Iteratively Quantized Image Retrieval Method

针对网络图像数据的迅速增多导致传统图像检索的效果不能满足当前需求的问题,提出了一种基于深度置信网络(deep belief network,DBN)和迭代量化(iterative quantization,ITQ)的无监督学习图像检索的方法.首先,构造深度置信网络的模型,此模型是由3层受限玻尔兹曼机堆叠而成;然后,用此深度置信网络模型对原始图像的高维特征进行中维特征提取,再采用迭代量化的哈希方法,对提取图像中维特征进行二值编码;最后,针对MNIST、CIFAR-10和Corel-1000数据集对模型进行实验验证并评估.结果表明,所提出的方法与现在的几种主流方法相比检索性能更好.除此之外,本方法对乳腺数据集DDSM和肺结节CT图像数据集LIDC-IDRI中的检索也取得了较好的效果.

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.