Datasets and reconstruction code for a virtual wave non-line-of-sight imaging approach
datasetposted on 02.08.2019, 13:05 by Xiaochun Liu, Ibon Guillen, Marco La Manna, Ji Hyun Nam, Syed Azer Reza, Toan Huu Le, Adrian Jarabo, Diego Gutierrez, Andreas Velten
This dataset contains the experimental data and code needed to replicate the non-line-of-sight (NLOS) imaging technique detailed in the accompanying paper "Non-line-of-sight imaging using phasor-field virtual wave optics".
NLOS imaging permits the obseration of objects partially or completely blocked from direct view by analyzing indirect reflections from the subject, reflected of a secondary relay surface. Where existing NLOS methods are limited in their application to simple, carefully controlled cases, the novel method presented here and in the accompanying paper permits imaging of complex scenes affected by confounding factors such as ambient light or a large depth range in the subject scene.
This dataset, contained in DataCode_Phasor_Field_VWNLOS.zip, consists of 5 distinct parts, each of which is described below. All code and data files are compatible with MATLAB.
A confocal non-line-of-sight image (CNLOS) reconstruction method used for comparison with the novel virtual wave method, and code to reproduce the comparison presented in the accompanying paper. The data and CNLOS method are originally described and provided in O’Toole, M., Lindell, D. B., & Wetzstein, G. Confocal non-line-of-sight imaging based on the light-cone transform. Nature 555, 338 (2018).
Experimental and virtual datasets used for analysis.
Code to reconstruct the geometry of all scenes using a filtered backprojection (laplacian filter, laplacian of gaussian filter). The backprojection C++ solver is implemented as two thread versions including Windows (matlab -R2018) and Unix.
Corresponding MATLAB scripts are: main_laplacian.m (for Windows system), main_laplacian_unix.m (for unix system)
Main Phasor Fields
Code to reconstruct the geometry of all scenes using our algorithm with a backprojection solver. The backprojection C++ solver is implemented as two thread versions including Windows (matlab -R2018) and Unix.
Corresponding MATLAB scripts are: main_phasor_field.m (for Windows system), main_phasor_field_unix.m (for unix system)
The NLOS letter refoucsing example using the fast Fresnel Diffraction approximation solver and exact Rayleigh-Sommerfeld Diffraction solver (RSD discrete integration implementation), illustrated in Figure 4 of the accompanying paper.