Four-direction weighted least squares phase unwrapping iterations algorithm is proposed based on regional mask filtering and the method of quality map leading weigh to solve the problem that interferometric synthetic aperture radar interferogram is difficult for phase unwrapping due to local noise, shadows or stripe breaks, which result in residual focus and unreliable data patches. Firstly, the noise patch region is masked and filtered. Therefore the real phase information is preserved and the error propagation is prevented effectively. Then, the four-direction least-squares iteration method, which can redefine the weight value by using the quality map method which can describe the maximum phase gradient and the change of phase derivative, is used to fill the mask region to suppress the error propagation, increase the computing speed and improve the over-smoothing effect. Simulation and real data experiments show that the proposed method has high unwrapping accuracy and can restore the original phase of the local patch of interferometric synthetic aperture radar interferogram. Compared with the traditional least squares method, it has the advantages of larger bandwidth of the difference equation matrix, better filling effect, less number of iterations and higher accuracy, which can solve the difficulty of interferometric synthetic aperture radar image unwrapping caused by noise concentration.