Low Photon Count Phase Retrieval Using Deep Learning

TL;DR

This paper demonstrates that deep neural networks can effectively perform phase retrieval at extremely low photon counts, outperforming classical algorithms and enabling object recovery with as little as one photon per pixel.

Contribution

The study introduces a deep learning approach for phase retrieval under low-light conditions, achieving successful object reconstruction with minimal photon counts and improving results with initial estimates.

Findings

Deep neural networks outperform classical algorithms at low photon counts.

Objects can be reconstructed with as little as one photon per pixel.

Training with an initial object estimate enhances phase reconstruction.

Abstract

Imaging systems' performance at low light intensity is affected by shot noise, which becomes increasingly strong as the power of the light source decreases. In this paper we experimentally demonstrate the use of deep neural networks to recover objects illuminated with weak light and demonstrate better performance than with the classical Gerchberg-Saxton phase retrieval algorithm for equivalent signal over noise ratio. Prior knowledge about the object is implicitly contained in the training data set and feature detection is possible for a signal over noise ratio close to one. We apply this principle to a phase retrieval problem and show successful recovery of the object's most salient features with as little as one photon per detector pixel on average in the illumination beam. We also show that the phase reconstruction is significantly improved by training the neural network with an initial estimate of the object, as opposed as training it with the raw intensity measurement.