基于Hadoop的图像纹理特征提取  

Image Texture Feature Extraction Based on Hadoop

作　　者：赵进超[1] 朱颢东[1] 申圳[1] 李红婵[1] 

机构地区：[1]郑州轻工业学院计算机与通信工程学院,河南郑州450002

出　　处：《郑州大学学报（工学版）》2015年第4期109-113,共5页Journal of Zhengzhou University（Engineering Science）

基　　金：国家自然科学基金资助项目(61201447);河南省高等学校青年骨干教师资助计划项目(2014GGJS-084);河南省科技创新杰出人才计划项目(134200510025);河南省教育厅科学技术研究重点项目(13A520367);郑州轻工业学院校级青年骨干教师培养对象资助计划项目(XGGJS02)

摘　　要：随着数字图像规模的不断增加,图像纹理特征提取已成为制约数字图像处理性能的一个关键步骤.Hadoop是一个性能卓越的开源大数据处理云平台,其向用户提供了MapReduce,HDFS等模块.首先对Hadoop平台、编程框架和Tamura纹理特征进行了介绍,然后将图像纹理特征提取过程在Hadoop平台上进行了实现.在这个过程中,每个Map任务对应一个图像文件,各节点可以同时提取集合内图像的纹理特征.实验表明:在图像数量较少和分辨率较低的情况,Hadoop不同节点数量所用时间并无太大差异.在图像分辨率较高且数量较多的情况下,Hadoop平台表现出较高的计算效率.With the increasing amount of digital image data,image texture feature extraction has become a key step of digital image processing. As an excellent massive data processing and storage capacity of the open source cloud platform,Hadoop provides a parallel computation model MapReduce,HDFS distributed file system module. In this paper,we firstly introduced Hadoop platform programming framework and Tamura texture features. And then,the image texture feature extraction was carried out on the Hadoop platform. In the process,every Map task corresponds to an image file,every nodes work at same time. The comparison results show that number of nodes have no influence about the processing time,when we have little images and the image has low-resolution. On the contrary,Hadoop paltform is more effective.

关 键 词：HADOOP Tamura纹理特征 图像处理 特征提取 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]