The phase unwrapping procedure is the most important step in three-dimensional(3-D) surface reconstruction. Among numerous techniques of phase unwrapping, the region-based algorithms are outstanding, which provide a trade-off between robustness and the computational complexity. However, there is indeed a serious weakness in the region-based algorithms that the unwrapped data may be incorrect when there are sudden appearances or high noise on measured objects’ surfaces. To overcome this problem, we propose a new phase-unwrapping algorithm by calculating the quality map of wrapped phase first. A a new region divided and coalescent strategy is also made with the help of second order difference quality map. Experimental results have demonstrated that the proposed algorithm is reliable and robust; it can improve the precision of wrapped data with sudden appearances or high noise remarkably.