Single-photon CASSI: towards ultralow-light spectral imaging.

Area of Science:

• Optics and Photonics
• Image Processing
• Computational Imaging

Background:

• Spectral imaging is crucial for various applications but is limited in ultralow-light conditions.
• Existing coded aperture snapshot spectral imaging (CASSI) systems face challenges with sparse spatial sampling and lack of specialized networks for low-light data.

Purpose of the Study:

• To develop a highly sensitive snapshot spectral imaging system for ultralow-light conditions.
• To integrate single-photon avalanche diodes (SPAD) with the CASSI framework for improved spectral and spatial information capture.

Main Methods:

• Developed an efficient spectral-spatial model tailored for ultralow-light imaging.
• Created a SPAD-CASSI simulation algorithm incorporating calibrated photon counts, noise, and spectral-spatial encoding.
• Designed a deep learning network for spectral reconstruction from simulated SPAD-CASSI data.
• Simulated data from 450-650 nm across 8 channels at 64x32 resolution with 100 µs exposure at 0.15 lux.

Main Results:

• The proposed SPAD-CASSI system demonstrates significantly improved performance in snapshot spectral imaging under ultralow-light conditions.
• Experimental validation across macroscopic, microscopic, and high-speed scenarios confirmed the system's robustness.
• The deep learning approach effectively reconstructed spectral information from sparse, noisy data.

Conclusions:

• The integrated SPAD-CASSI system offers a robust solution for spectral imaging in challenging ultralow-light environments.
• This advancement opens new possibilities for applications requiring high sensitivity and spectral resolution in low-light settings.