Improving Content Based Video Retrieval Performance by Using Hadoop-MapReduce Model

Abstract

In this paper, we present a distributed Content-Based Video Retrieval (CBVR) system based on MapReduce programming model. A CBVR system called bounded Coordinate of Motion Histogram (BCMH) has been implemented as case study by using Hadoop framework. Our work consists of proposing a distributed model to extract videos signatures and compute similarity with the BCMH system based on a set of Mapreduce jobs assigned to multiple nodes of the Hadoop cluster in order to reduce computation time of training process. The proposed approach is tested on HOLLYWOOD2 dataset and the obtained results demonstrate efficiency of the proposed approach.