A parallel algorithm for change detection

Abstract

Change detection is an important research area in image and video processing due to wide applications in medicine, surveillance, remote sensing, geographical information processing and in defense and security. This paper presents a real-time distributed algorithm to detect changes in videos. The proposed algorithm may be used in many change detection applications that require real-time performance like CCTV Security Surveillance Camera Systems. In such applications fast change detection is very important. Many a times, we need to search a video captured by security camera for an event with significant change. Doing it manually is very time consuming and cumbersome, a fast automatic algorithm is desired in such cases. The proposed parallel algorithm runs on a cluster and detects the significant changes in real-time. The algorithm is implemented in C using MPICH2 in Linux environment and is tested over a number of videos for correctness, accuracy and execution time. The proposed approach is also compared with serial change detection techniques and gained a speedup of approximately 3.4 on cluster with 4 average machines.