Image Texture Feature Extraction Based on Hadoop Cloud Platform and New ImageClass

摘要

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 in Hadoop platform. In the process, every image file is treated as a split which will be processed as a job record, every Map task corresponds to an image file, we use a Reduce task to write the result to the specified location in the specified format. Meanwhile, we defined an ImageClass which it not only can achieve the basic function modules of hadoop, but also can increase some modules based on the actual need. The comparison results show that for image texture feature extraction of low-resolution images, Matlab platform has more obvious advantage than Hadoop platform, but for image texture feature extraction of high-resolution images, the spent time in Hadoop platform is lesser, data processing capability of Hadoop platform is better.