Events Meet Phase-Shifting Digital Holography: Practical Acquisition, Theory, and Algorithms

TL;DR

This work advances phase-shifting digital holography by introducing FE-PSDH, which leverages a hybrid event-based vision sensor to phase-shift during a single exposure. It provides a closed-form, noiseless reconstruction and an optimization-based approach to stabilize reconstructions in real data, using the EVS to simultaneously capture a blurred hologram and event streams. Through simulations and optical experiments, FE-PSDH demonstrates reconstruction quality comparable to conventional F-PSDH while dramatically reducing acquisition time, thanks to single-exposure phase shifting. The method promises practical gains in holographic imaging by enhancing acquisition efficiency without sacrificing wavefront fidelity.

Abstract

We introduce a novel phase-shifting digital holography (PSDH) method leveraging a hybrid event-based vision sensor (EVS). The key idea of our method is the phase shift during a single exposure. The hybrid EVS records a hologram blurred by the phase shift, together with the events corresponding to blur variations. We present analytical and optimization-based methods that theoretically support the reconstruction of full-complex wavefronts from the blurred hologram and events. The experimental results demonstrate that our method achieves a reconstruction quality comparable to that of a conventional PSDH method while enhancing the acquisition efficiency.

Figures (9)

• Figure 1: Phase-shifting digital holography (PSDH). Red region: optical path. BS: beam splitter. M: mirror. PS: phase shifter. IF: interference fringe. $k$: wavenumber. $\phi$: phase-shifting value. Conventional method: ordinary image sensor. Our method: hybrid event-based vision sensor (EVS).
• Figure 2: Timing charts of acquisition. $t_\mathrm{sens}$: exposure time of sensor. $t_\mathrm{shift}$: transient time of phase shifter. Gray bar: total acquisition time. Phase-shifting value is initialized to $0$ before the first exposure. Throughout this paper, we assume $t_\mathrm{shift} \leq t_\mathrm{sens}$ and both methods share $t_\mathrm{sens}$ and $t_\mathrm{shift}$. Acquisition time of conventional method: $3t_\mathrm{sens}+2t_\mathrm{shift}$. Our method: $t_\mathrm{sens}$---single exposure time alone.
• Figure 3: Optical implementation of acquisition system in Fig. \ref{['fig:psdh']}. CL: collimator lens. SLM: spatial light modulator. ND: neutral-density filter. Ball and Pins are printed on OHP sheets. EVS--Ball: $120$ mm. EVS--Pins: $270$ mm.
• Figure 4: Errors by Algorithm \ref{['alg:analytical']} against $t_{\pi/2,\mathrm{anal}}$.
• Figure 5: Errors by Algorithm \ref{['alg:optimization']} against $\lambda$.