基于Hadoop的海量医学图像检索系统

Massive medical image retrieval system based on Hadoop

为了提高海量医学图像检索效率,针对单节点医学图像检索系统的缺陷,提出一种基于Hadoop的海量医学图像检索系统。首先采用Brushlet变换和局部二值模式算法提取医学示例图像特征,并将图像特征库存储于Hadoop分布式文件系统(HDFS);然后采用Map将示例图像特征与特征库的特征进行匹配,采用Reduce接收各Map任务的计算结果,并按相似度大小进行排序;最后根据排序结果找到医学图像的最优检索结果。实验结果表明,相对于其他医学图像检索系统,Hadoop的医学图像检索系统减少了图像存储和检索时间,提高了图像检索速度。

In order to improve the retrieval efficiency of massive medical images, a new medical image retrieval system was proposed based on distributed Hadoop to solve the low efficiency of medical image retrieval system based on single node. Firstly, the features of medical image were extracted by using Brushlet transform and Local Binary Pattern （LBP） algorithm, and the feature database was stored in the Hadoop Distributed File System （ HDFS）. Secondly, the Map was used to match the features of retrieval images and medical images in the library, and the matching results of the Map task were collected and sorted by the Reduce function. Finally, the optimum results of medical image retrieval were obtained according to the ordering. The test results show that, compared with other medical image retrieval systems, the proposed system reduces the time of image storage and retrieval, and improves the image retrieval speed.