Hadoop平台下海量图像处理实现

摘要

For that Hadoop does not directly support to deal with the problems of massive small image file and efficient management of these small files,and to let developers who do not understand the internal mechanism of Hadoop develop and deal with the massive image algorithm,the Hadoop framework for image processing was researched and implemented.Framework for the development of image data in the Java development language was designed,so that it could be applied to the MapReduce computing model.A storage model was designed to efficiently store massive small files and manage small files.To make developing easy,within the processing framework for massive images,an outwardly programming interface adapting to a variety of image processing algorithms was provided.This framework allowed developers of image processing applications to leverage the Hadoop MapReduce framework without having to master its technical details.%针对Hadoop不直接支持处理海量小图像文件且不能高效管理这些小文件的问题,为让开发者在不了解Hadoop内部机制的情况下开发处理海量图像的算法,研究并实现Hadoop图像处理框架.制定图像数据在Java开发语言中的表示,使其适用于MapReduce计算模型;设计存储模型高效存储海量小文件及管理小文件;为让开发变得简单和高效,在框架中,向外提供一个适用于各种图像处理算法的编程接口,使图像处理开发人员在不用了解Hadoop内部复杂细节的情况下使用这个框架在海量图像处理中发挥杠杆作用.实验结果表明,该框架能高效处理海量图像,有很强的适用性.