Abstract
We propose a novel method to reconstruct non-line-of-sight (NLOS) scenes that combines polarization and time-of-flight light transport measurements. Unpolarized NLOS imaging methods reconstruct objects hidden around corners by inverting time-gated indirect light paths measured at a visible relay surface, but fail to reconstruct scene features depending on their position and orientation with respect to such surface. We address this limitation (known as the missing cone problem) by capturing the polarization state of light in time-gated imaging systems at picosecond time resolution, and introducing a novel inversion method that leverages directionality information of polarized measurements to reduce directional ambiguities in the reconstruction. Our method is capable of imaging features of hidden surfaces inside the missing cone space of state-of-the-art NLOS methods, yielding fine reconstruction details even when using a fraction of measured points on the relay surface. We demonstrate the benefits of our method in both simulated and experimental scenarios.
Overview
Schematic comparing an unpolarized NLOS method and our method leveraging directional information encoded in the polarization. (a) Unpolarized NLOS light transport model depicted as a confocal setup for simplicity. (a, left) The light emitted by laser π travels along path β¨π β π₯π β π₯π£ β π₯π β π β© before reaching detector π. Both laser π and detector π are coaxial and aim at the same point of the relay surface. (a, right) When inverting a model in unpolarized methods, the only information of the measured light is its time of flight and, hence, all the possible candidates π₯π£ lie on a hemisphere with radius πΞπ‘π£ /2. As a result, there is no definite recoverable direction. (b) Our polarized NLOS transport model, also depicted as a confocal setup for simplicity. (b, left) We analyze the polarizing effects of a conductive, micro-faceted relay wall to retrieve directional information. (b, right) If the hidden object is depolarizing, we invert the polarizing effect of the last bounce to determine the direction of the ray and, hence, recover the direction to π₯π£ . Note that the analyzer is only placed in front of the detector.
Results
Simulated captures & reconstruction
Simulated results comparing reconstructions of Bunny and Lucy. Previous works (LCT [OβToole et al. 2018], f-k migration [Lindell et al. 2019], phasor fields [Liu et al. 2019b]) fail to reconstruct certain features when they fall into the missing cone (e.g., Bunnyβs ears, Lucyβs wings and base). By leveraging polarization, our method yields full reconstructions even when using 16 times less scanned points (without decrease in quality, but reducing reconstruction time). As shown in (c), our method also produces high-quality reconstructions in the presence of polarizing rough conductors materials for the hidden object.
Real captures & reconstruction
(a) Our prototype NLOS setup. For our results, we capture only one point on the relay surface. (b) Top-view schematic diagram. We use aluminum for the relay surface, and tested two different materials for the target (hidden object): depolarizing drywall, and polarizing aluminum. (c) Reconstruction results (top view) for the target at 0 and 90 degrees, the latter falling entirely in the missing cone. We compare against simulated results that closely match the actual capture conditions. Note that previous methods can not reconstruct neither at 0 nor 90 degrees, since a single-point capture is limited to the directional ambiguities present in conventional NLOS methods (see Figure 2).
Paper
Bibtex
Related Work
- 2022: Non-Line-of-Sight Transient Rendering
- 2021: Polarimetric spatio-temporal light transport probing
- 2019: Analysis of Feature Visibility in Non-Line-of-Sight Measurements
Acknowledgements
We thank the anonymous reviewers for their time and insightful comments, aswell as the members of the Graphics and Imaging Lab for their help with the manuscript. Our work was funded by the European Union's Defense Fund Program through the ENLIGHTEN project under grant agreement No. 101103242, and Oscar Pueyo-Ciutad was also partially supported by a FPU22/02432 predoctoral grant from the Ministerio de Ciencia, InnovaciΓ³n y Universidades.