Novel algorithms for phase retrieval

Many phase retrieval algorithms are used for the designing process of diffractive optical element (DOE). To design a DOE, it is almost the occasions to recovery phase distribution on the micro-optics element when the amplitude distribution of inputting plane and outputting plane has been known. Generally, there are two kinds of strategies to resolve phase retrieval problems, namely iterative method and try-and error method. There are a lot of famous algorithms have been successful used in this field, such as Gerchberg and Saxton (GS), ST, Input-Output, Yang and Gu (YG) etc. In this paper, a novel algorithm is brought forward. The phase on DOE is considered as the superposition of a serial of sinusoidal curves. Randomfly, a concrete sinusoidal phase disturbance with definite frequency and amplitude is selected. Correspondent disturbance complex amplitude distribution on image plane can be considered as transformation result of the disturbance phase by the optical system. Then, new complex amplitude on image plane can be calculated as the convolution of its original one and the disturbance one. By Simulated Annealing algorithm, the optimal frequency and amplitude for the sinusoidal phase disturbance are searched. Gradually, when enough sinusoidal phase disturbances are searched, the phase of DOE will approximate to its global optimum solution. At last, for the optimum phase solution, phase smoothing method is used to relieve jump discontinuity.