A New Change Detection Algorithm for Visual Surveillance System

This paper describes the implementation of a background subtraction approach for indoor surveillance applications and presents a novel change detection algorithm. This algorithm uses a color similarity metric that considers the information of magnitude and phase of the difference between the background model and each video frame, both represented in the RGB color space. One of the advantages of such combination is that both features, magnitude and phase, take into account the intensity and the chromaticity information from a color image. The approach was tested using the PETS2004 database obtaining acceptable results.