10 October 2025 in Optics Letters
Non-line-of-sight (NLOS) imaging seeks to reconstruct scenes obscured from the direct line of sight by capturing indirect light encoded on an intermediate visible surface. Unlike active approaches that rely on expensive and dedicated systems, passive methods utilize only a low-cost digital camera to record the light signal encoded on visible surface. However, the imaging range of passive NLOS techniques is typically constrained to the meter scale due to a low signal-to-background ratio (SBR), resulting from significant light attenuation over long propagation distances and the presence of strong ambient background illumination. In this work, we present an approach that combines pattern-based calibration with low-rank matrix decomposition for noise suppression, enabling both enhancement of the NLOS signal and separation of ambient background contributions. We experimentally demonstrate passive NLOS imaging at a 10 m wall-to-camera distance and a 4.5 m wall-to-object distance, achieving centimeter-scale resolution at an SBR as low as –13 dB, outperforming previous works. Our results highlight the significant potential for practical passive NLOS applications.
29 June 2026 in ACM Transactions on Graphics
Thermal non-line-of-sight (NLOS) imaging utilizes the thermal radiation emitted by hidden objects to “see around corners”, offering promising applications in security and search-and-rescue missions. However, reconstructing hidden scenes through rough relay walls remains highly challenging due to severe spatial mixing caused by complex reflection and scattering. While existing model-based and learning-based approaches yield high-quality results under reflective surfaces, they struggle to deliver satisfactory performance under rough surfaces. Here we introduce NLOSFormer, a physics-embedded neural network to address this challenge. By formulating light transport as a convolution process, NLOSFormer estimates the convolution kernel from measurements to guide reconstruction within an end-to-end network. This explicitly estimated kernel effectively narrows the solution space, significantly improving generalization and enabling relative depth estimation. To support training and evaluation, we further present a publicly available dataset named ThermalNLOS, together with a novel data augmentation strategy to alleviate overfitting. Extensive experiments demonstrate that NLOSFormer significantly outperforms existing approaches across a wide range of relay walls. Notably, it achieves real-time imaging of dynamic targets under rough surfaces, representing a substantial advance toward robust and practical deployment.
25 August 2026 in ACM Transactions on Graphics
This is an erratum for the article “Thermal Non-Line-of-Sight Imaging through Rough Surfaces” published in ACM Trans. Graph. 45, 5, Article 41 (June 2026), 21 pages.