Online figure-ground segmentation with adaptive metrics in Generalized LVQ