Abstract

We propose an ADMM-based phase retrieval algorithm for FP that utilizes anisotropic total variation regularization for the object function and L2 regularization for pupil functions. All our results in simulation and real experiments demonstrate that our algorithm outperforms the Gauss-Newton algorithm in terms of object and pupil function recovery. Our findings suggest that our algorithm enables the reconstruction of the objective function using a shorter exposure time and fewer measurements as a factor of 60x.

© 2023 The Author(s)