Abstract: The existing low-rank matrix completion algorithms are specially designed for single-type mask removal, but they difficultly adapt to the challenging situation with multiple mask types.