The phase shifting interference technology is one of the principal methods of high accuracy surface profile measurement. The phase unwrapping is a challenging problem when the pattern contains high level noises and bulk invalid regions. We improve seed-unwrapping algorithm through identifying the residues with combination of their three neighborhoods, marking discontinuities before unwrapping and after unwrapping, and recovering the residues and discontinuities with average value of their neighborhoods after unwrapping. The simulation indicates that it is robust for unwrapping of patterns with high noises and bulk invalid regions, and there are no high noisy points in the images unwrapped by the improved algorithm, and the unwrapped images can be well smoothed by average value filter.