A review on the development of 3D image reconstruction algorithms based on sparse single-photon data

Single-photon light detection and ranging (SP-LiDAR), which is recognized for its single-photon sensitivity and picosecond-level time resolution, excels at extracting target information from weak signals by accumulating multiple counts. This technology has been extensively applied in precise cartographic mapping and accurate navigation of autonomous vehicles. Owing to the advantages of image reconstruction algorithms in single-photon high-resolution real-time imaging, including low cost, minimal technical complexity, and superior reconstruction quality, these algorithms have become a focal point in single-photon imaging research. To address the challenges of sparse signals and intense noise in single-photon imaging, researchers have employed regularized optimization algorithms, Bayesian probability models and deep learning architectures to extract target features under adverse conditions, significantly enhancing the performance of imaging systems. Based on the detection and imaging principles of single-photon LiDAR, this paper critically reviews typical research on image reconstruction algorithms in photon-starved regimes and their advancements for complex detection targets and attenuative transmission media and discusses potential future directions for these algorithms.