Automatic segmentation of moving objects in video sequences: a region labeling approach