Classification of change detection algorithms for object-based applications

Change detection is a temporal segmentation tool aiming at identifying changes in image sets or image sequences at two different times. Many change detection algorithms have been proposed over the past decade for the generation of video objects in a wide range of applications, ranging from interactive multimedia to remote surveillance. Most of these algorithms are tailored to the specific application at hand. Therefore there is a need of a general model for change detection which could support the choice of the optimal change detection strategy. We present a unified approach to change detection based on a 4-step model. Change detection algorithms proposed in the literature are analysed in terms of the four building blocks of the proposed model. This study is useful not only for reviewing the state of the art of change detection, but also for refining and improving change detection methods, by providing useful guidelines for the better use of the algorithms.