Adjustable frequency filtering and weighted feedback for iterative phase retrieval under noisy conditions

Abstract

In coherent diffraction imaging (CDI) systems, the results of phase retrieval could be deteriorated by the inevitable experimental noise. To address the problem, a noise-robust algorithm is designed with weighted feedback and adjustable frequency filtering. The frequency filter serves as the smoothness constraint and has an adaptive parameter during iterations, to suppress the noise as well as avoid the convergence stagnancy of iterative phase retrieval. Numerical simulations and experimental results demonstrate that the proposed algorithm is superior in reconstruction quality under noisy conditions and increases the convergence speed. Furthermore, the minimum acceptable number of measurements of diffracting images is reduced. This work is beneficial to data synthesis in CDI systems.