并行多媒体数据库中基于内容的高效检索的数据分配方法的研究

摘要

Zoom in to a pro-rata basis with high-dimensional metadata describing the characteristics of large-scale multimedia database, and can provide content-based fast retrieval (CBR) is becoming increasingly important in many applications. To achieve this goal, the general parallel, shared-nothing architecture implementation. But data between different nodes to be how to allocate in order to obtain efficient parallel content-based retrieval results. By assuming a clustering process based on the complexity of the analysis of CBR, an optimal number of clusters and number of nodes as the characteristics of the data distribution method. And K nearest neighbor query processing algorithms implemented in the different aspects of high-dimensional synthetic database experiments to verify.%一种可按比例放大到带高维元数据描述特征的大型多媒体数据库,并且能提供基于内容的快速检索(CBR)的方法在许多应用领域变得越来越重要.为实现这个目标,通过通用的并行无共享架构实现.通过假设一个聚类过程并根据对CBR的复杂程度分析,提出了一种以最佳聚类数和节点数为特点的数据分配方法.通过在不同的高维数合成数据库方面进行实验对其进行了验证,并实现了对全K最邻近查询处理算法.