阐述了GS算法用于相位恢复的原理，并对算法的应用范围进行了扩展，不再局限于已知光瞳面和焦平面光强的情况。仿真中，选取了光束自由传输的两个垂轴截面，利用光强分布，采用了GS迭代算法对常见的几种波前畸变进行了恢复。针对相位恢复中GS算法收敛速度慢和精度低的问题，提出了迭代过程中，在光束输入端和输出端分别进行加权改进，给出了加权系数取值范围。仿真结果表明，这种改进有利于算法跳出局部极值，在加权值取1.2时，针对象散像差的恢复中，速度和精度提高了2倍左右，对一些其它常见像差的恢复精度也有所提高。

In this paper, the GS algorithm used for phase retrieval is elaborated, and the application scope of this algorithm is expanded, no longer confined to the known pupil surface and the focal plane intensity. In the simulation, two vertical axial section of the beam are chosen, where the optical power is detected to retrieve several kinds of common phase aberration based on GS algorithm. An improved GS algorithm is brought forward, which adds two weight coefficients in the circulation. The range of the two weight coefficients is given, too. Results of the simulation show that both the speed and precision of GS algorithm are improved.