Abstract
Processing video data is becoming more useful in a wide range of applications. However, video data are demanding for computing resources, such as processor, memory, and disk. This is because the data size is huge in nature and growing exponentially. Change detection is a commonly used method in a variety of video processing applications, so it has been attracting the attention of many researchers. The goal of improving the speed of change detection could be to satisfy real-time performance or to process larger data in a timely manner. This study proposes an approach based on MapReduce and sampling to improve the performance of using change detection to process large video data on Hadoop clusters. The experiments, conducted on an outdoor scene dataset, show significant improvement in the execution time.