Content area
Full text
ABSTRACT
With recent improvements in camera performance and the spread of low-priced and lightweight video cameras, a large amount of video data is generated, and stored in database form. At the same time, there are limits on what can be done to improve the performance of single computers to make them able to process large-scale information, such as in video analysis. Therefore, an important research topic is how to perform parallel distributed processing of a video database by using the computational resource in a cloud environment. At present, the Apache Hadoop distribution for open-source cloud computing is available from MapReduce1. In the present study, we report our results on an evaluation of performance, which remains a problem for video processing in distributed environments, and on parallel experiments using MapReduce on Hadoop2.
Keywords: Hadoop, MapReduce, Image Processing, Sequential Image Database
1. INTRODUCTION
With the spread of cloud computing and network techniques and equipment in recent years, a large amount of data collected in a variety of social situations have been accumulated, and so the need for analysis techniques to take advantage of useful information that can be extracted from such data sets is increasing. This is also true for video data, in which images are sequential and the data includes the associated time and frame information of each frame. Today, because video cameras are set up to perform surveillance of moving objects such as pedestrians and vehicles, a large amount of video data is generated, and stored in database form. To improve these video database systems, which hold a large amount of video and related data, parallel processing using CPUs and disks at multiple sites is an important area of research.
Also, when considering operations such as search and other types of analysis of video images recorded by video camera and stored in a database, there are limits on what can be done to improve the performance of single computers to make them able to process large-scale information. Therefore, the advantages of parallel distributed processing of a video database by using the computational resources of a cloud computing environment should be considered. In addition, if computational resources can be secured easily and relatively inexpensively, then cloud computing is suitable for handling large video databases at...