Visual surveillance aims to reliably extract foreground objects. Traditional algorithms usually use a background model image which is generated through the probabilistic modeling of changes over time ...