What are some good algorithms to convert motion vector from 4*4 block/grid level to pixel level considering motion discontinuity at boundaries and occlusions as well for optical
