Design And Implementation Of Distributed Image Computing And Retrieval System

The development of multimedia technologies has brought great changes of people’s lives. It not only provides basic life convenient, but is also changing the traditional way of life, entertainment and communication. However, with the advent of the information age, the amount of data expansion and the requirements for higher speed of information processing from users, multimedia technologies have been processed in some way bottlenecks.As an important part of cloud computing, distributed computing is wide spread in the industry. As a general idea for distributed computing platform, MapReduce proposed by Google has already been applied to many fields, especially for mass of data storage and computing. The calculated delay is shortened through breaking tasks down and processing parallel, and it improves processing efficiency significantly.This thesis mainly processes on the image information using some technologies based on distributed computing, and designs and implements a distributed image computing and retrieval system. The system mainly consists of two functions including image computing and image retrieval. The function of image computing includes extracting feature points, training images and generating vocabulary tree etcetera. The main goal of this system is to satisfy the users’ requests of image retrieval, and users can obtain augmented reality information related to the image after system processing.The implementation of this system draws on the general framework and principle of MapReduce, some research is done on task scheduling, and reasonable task scheduling strategies are put forward to make the system have a high efficiency. Meanwhile, I focus on the functions associated with image computing and retrieval, and propose the corresponding distributed solutions. I also take a little consideration of how the data is stored, and give methods to speed up the serialized data access.The correctness of the system functionality is confirmed through the testing of the system and the experimental data, and performance indicator has reached the height as it required.