Scene interpretation by fusion of segment and region information