Hybrid phase retrieval algorithm based on modified very fast simulated annealing

Abstract

AbstractThe Misell algorithm is one of the most widely used phase retrieval holography methods for large reflector antennas to measure surface deformation. However, it usually locks in a local minimum because it heads downhill from an initial estimation without any consideration whether it heads for a global minimum or not. The core problem of the Misell algorithm is to find an initial estimation near the global minimum to avoid local stagnation. To cope with the problem, we construct a hybrid Misell algorithm, named modified very fast simulated annealing (MVFSA)-Misell algorithm, to search for the global minimum with a high efficiency. The algorithm is based on the combination of the MVFSA algorithm and Misell algorithm. Firstly, the MVFSA is utilized to obtain a rough position near the global minimum in limited steps. Then, the Misell algorithm starts from the rough position to converge to the global minimum with high speed and accuracy. The convergence characteristic of the proposed algorithm was discussed in detail through digital simulation. Simulation results show that the algorithm can reach global minimum in a very short time. Unlike the traditional Misell algorithm, the hybrid algorithm is not influenced by initial phase estimation.