We only briefly outline the techniques we use here, since they are standard computer vision techniques and more appropriatly described in papers such as [11,12,5,13]