Abstract
Single-photon imaging provides high photon sensitivity and the capability to capture ultrafast dynamics. However, temporal cutoff characteristics in single-photon avalanche diode (SPAD) arrays arise from in-frame dead time caused by the avalanche process and inter-frame dead time caused by the readout circuit, limiting the achievable frame rate when exposure time is reduced. We first studied a physics-based temporal model that introduces in-frame and inter-frame dead time, and proposed two reconstruction strategies that achieve higher fidelity and temporal resolution. Then we designed a transformer network with temporal and spatial feature extractors, which achieved 2× temporal resolution, 2× spatial resolution, and average peak signal-to-noise ratio improvement of 8.14 dB. We applied the technique to a series of observation experiments, including fan rotation, plasma discharge, and fluorescence quenching dynamics. These experiments validate the technique’s state-of-the-art temporal and spatial super-resolution SPAD imaging performance.
| Original language | English |
|---|---|
| Pages (from-to) | 3121-3130 |
| Number of pages | 10 |
| Journal | Photonics Research |
| Volume | 13 |
| Issue number | 11 |
| DOIs | |
| Publication status | Published - Jan 2025 |