The Study Of Massive Image Processing Based On Cloud Computing

With the progress in information technology and the rapid development of image data acquisition technology, large amounts of multimedia data which is dominated by digital image are produced in various walks of life daily. Being confronted with the explosive growth of digital image data, the traditional stand-alone mode for image processing exposes out many problems:such as low processing speed, poor concurrency, etc. As the traditional mode is gradually unable to satisfy customers’ require, therefore seeking a new efficient image processing mode is quite necessary in nowadays. Cloud computing is a new model of computing, which is a typical distributed, parallel computing that can greatly reduce the execution time of computing tasks. As speeding up data processing is urgently needed, more and more image processing algorithms are implemented on cloud computing platform for parallel distributed computing.Hadoop platform is an open source cloud computing platform, which provides MapReduce computing model, HDFS distributed file system, etc. This paper presents a parallel massive image processing model which is based on Hadoop platform. And Hadoop Streaming technology is used to design and implement large-scale image processing parallelization. Furthermore Breast X-ray image processing is cited as an example of the experiments. During the experiment, a shell script is applied as an interface. Then the calling process of image processing program is written on this script, which ultimately simplify the modification for image processing program moreover is easily applied to a variety of image processing algorithms. Based on the distributed storage and computing of Hadoop platform there is an advantage of high reliability and scalability for large-scale parallel image processing. Thus the purpose of rapid processing of large-scale images is easily realized. This paper also studies the task allocation strategy for large-number image processing on the isomorphic Hadoop cluster and heterogeneous Hadoop cluster. For the heterogeneous cluster, a task allocation strategy based on GA (Genetic Algorithms) is used for optimizing the allocation of large-number images processing. Experimental results reveal that the optimization can significantly speed up the image processing.