Under-sampled single-photon non-line-of-sight imaging reconstruction based on spatio-temporal correlation

Non-line-of-sight imaging captures information from hidden objects by projecting or scattering their signals onto an intermediate surface, enabling target observation beyond line-of-sight obstacles. Among various approaches, pulsed lasers combined with single-photon detectors have become widely adopted in recent years. However, due to hardware limitations, the data acquisition process requires point-by-point scanning of the hidden scene to obtain complete target information. Moreover, each scanning position necessitates a certain integration time to accumulate sufficient photon counts for highquality image reconstruction. As a result, the overall acquisition process is time-consuming. Recent efforts to reduce integration time have improved acquisition efficiency but at the cost of degraded image quality, such as blurred contours and lower signal-to-noise ratios, which impair reconstruction performance. To address these challenges, this paper proposes a reconstruction method that leverages spatio-temporal correlations of photon counts based on wave propagation model. A spatial enhancement module and a temporal constraint module are introduced for data resampling and image reconstruction. The spatial enhancement module strengthens target responses and suppresses noise by analyzing correlations between adjacent sampling points, while the temporal constraint module combines the temporal correlation of photon counts with distance-based attenuation to enhance structural reconstruction and further suppress noise. Experiments on public datasets show that the proposed method significantly enhances signal-to-noise ratio and structural similarity under reduced integration times, outperforming traditional methods like f-k migration.