基于Hadoop分布式计算的图像检索方法

Image Retrieval Based on Hadoop Distributed Computing

针对传统单节点的基于内容的图像检索方法计算速度较慢,检索效率不高的问题,提出了一种基于Hadoop分布式计算的图像检索方法。首先提取出图像的颜色、纹理和形状特征用于表示图像,在检索阶段将检索任务分配到各个Map子节点,所有的Map结果根据相似度非减进行排序,并将前N个结果进行输出。实验结果表明,该方法有效地利用了云计算平台的并行处理能力,相比较单节点的图像检索方法,提高了CBIR方法的运行效率。

Aiming at the problem of inefficiently of traditional content based image retrieval method, this paper proposes an imageretrieval method based on hadoop distributed computing. We first extract color features, texture features and sharp features from im-age dataset. Then, the retrieval job will be distributed to all map services and all retrieval results will be sorted according to match-ing results. Finally, the first N images will be sent back to user. The experimental result shows that our proposed method can im-prove the computing efficiency of traditional CBIR method.