Answers are grounded in this paper and cite their sources.
Non-line-of-sight (NLOS) imaging leverages ultrafast lasers and time-resolved detectors to acquire multiple scattered photons from a relay wall, enabling the reconstruction of hidden scenes around the corner. However, the time-consuming dense scanning process hinders its practical application. Under-scanning improves acquisition efficiency but results in the loss of transient information. This weakens the measurement constraints available for inversion and degrades the quality of subsequent reconstructions. This paper proposes a transient measurement recovery method for under-scanning NLOS imaging systems that exploits inherent spatiotemporal correlation priors to recover transient responses, thereby improving the stability of hidden-scene inversion. Specifically, a transient completion network is designed to recover dense transient measurements from sparse observations by modeling spatiotemporal correlations and local signal variations. The recovered transients are subsequently mapped to the three-dimensional spatial domain through a volume reconstruction network to estimate the geometry of hidden targets. Experiments on both synthetic and real-world measurement datasets show that the proposed method recovers transient measurements to support reliable inversion of hidden-scene geometry under sparse scanning grids of 32×32 and 16×16. Even when the total acquisition time is reduced to 0.25 s, the recovered transient measurements retain sufficient information to reconstruct the main contours of the hidden target. These results suggest that computational recovery of missing transients improves measurement completeness and strengthens inversion reliability while preserving the high acquisition efficiency. The proposed method provides a measurement enhancement strategy for sparse time-resolved photon acquisition systems.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: