Phase retrieval algorithm using edge point referencing

Abstract

In the past few decades, extensive research and efforts have been made for developing a phase retrieval iterative algorithm (PRA) for reconstructing a complex object from far-field intensity equivalently from the object autocorrelation. Since most of the existing PRA techniques employ a random initial guess, the reconstruction output sometimes changes in different trials leading to a non-deterministic output. Additionally, the output of such algorithm occasionally either shows non-convergence, needs a longer time to converge, or shows the twin-image problem. Due to these problems, PRA methods are unsuitable for cases where consecutive reconstructed outputs need to be compared. In this Letter, a novel, to the best of our knowledge, method is developed and discussed using edge point referencing (EPR). In the EPR scheme, in addition to illuminating a region of interest (ROI) of the complex object, a small area near the periphery of the complex object within the ROI is illuminated with an additional beam. Such illumination creates an imbalance in the autocorrelation that can be used to improve the initial guess for achieving unique deterministic output free from the aforementioned problems. Furthermore, by introducing the EPR, one can also achieve faster convergence. To support our theory, derivation, simulations, and experiment are performed and presented.