Adaptive Filtering and Sequence Estimation for High-Precision Single-Photon 3D Imaging

Abstract

Single-photon imaging offers a powerful computational approach for high-precision depth reconstruction with high photon efficiency. However, its accuracy is fundamentally limited by time-of-flight (TOF) misestimations stemming from system jitter, background noise, and detector dead time, which may collectively introduce significant depth measurement errors. Here, we present an advanced single-photon imaging technique that combines photon sequence estimation with adaptive Gaussian filtering to overcome these limitations. Coates’s estimator is first employed to suppress photon pile-up effects and accurately reconstruct the incident photon sequence. Subsequently, a Markov chain-based adaptive Gaussian filtering algorithm is applied to correct peak shifts in the reconstructed photon histogram. Experimental validation demonstrates that the proposed method achieves millimeter-level depth resolution (∼2–3 mm) across a measurement range of 6.5 m. Compared with other reconstruction methods, our approach delivers over 50% improvement in reconstruction accuracy, establishing it as a versatile solution for high-precision three-dimensional (3D) imaging in robotics, medical diagnostics, and remote sensing.