Accurate phase retrieval of complex point spread functions with deep residual neural networks

TL;DR

This paper presents a deep residual neural network that accurately and efficiently reconstructs phase information from intensity data for complex point spread functions, enabling rapid optical system analysis.

Contribution

The study introduces a deep residual neural network approach for phase retrieval of arbitrary PSFs with Zernike-type phase modulations, demonstrating high accuracy and speed.

Findings

Successfully retrieves Zernike coefficients from PSFs at different focal slices.

Achieves rapid phase reconstruction with high accuracy.

Applicable to complex optical systems with arbitrary PSFs.

Abstract

Phase retrieval, i.e. the reconstruction of phase information from intensity information, is a central problem in many optical systems. Here, we demonstrate that a deep residual neural net is able to quickly and accurately perform this task for arbitrary point spread functions (PSFs) formed by Zernike-type phase modulations. Five slices of the 3D PSF at different focal positions within a two micron range around the focus are sufficient to retrieve the first six orders of Zernike coefficients.