SPAD - constructing images based on individual photon readings

This project explores the potential of Single-Photon Avalanche Diode (SPAD) cameras as a groundbreaking imaging modality for low-light and high-speed scenarios. SPAD cameras, capable of detecting and timestamping individual photons, open new avenues in imaging dynamic scenes. We conducted characterization experiments to understand SPAD behavior under varying light levels and exposure times, revealing significant challenges with low signal-to-noise ratios (SNR) in low-lux conditions. Addressing this, we reviewed state-of-the-art algorithms leveraging photon timestamps, identifying the lack of effective denoising as a critical gap. By assuming local homogeneity across pixel grids, we developed a method to reduce noise, enabling improved flux reconstruction over time and clearer dynamic frame captures.

To maximize imaging capabilities, we are now investigating hybrid sensor systems, combining SPAD cameras with conventional sensors. This interplay leverages SPAD’s strengths in low-light and dynamic scenarios while utilizing the high bandwidth and cost efficiency of conventional sensors for high-flux environments, creating a versatile solution for diverse imaging challenges.