How Can Blind Source Separation Aid in Discriminating Against Unwanted Sources in Non-Line-of-Sight Imaging?
Passive Non-Line-of-Sight (NLoS) imaging aims to reconstruct a scene hidden from the camera’s field of view by observing indirect light scattered off of a wall. A drawback of many NLoS imaging methods is presence of clutter, or unwanted signals that impede reconstruction attempts. However, multispectral analysis has been shown to be a valuable tool in mitigating clutter and improving NLoS reconstructions. This presentation builds upon previous methods by applying a set of signal processing techniques called Blind-Source-Separation (BSS) to the multispectral clutter problem. Theoretical limits of BSS were first explored in computer-based NLoS imaging simulations. Experiments were then performed using an LCD screen to generate multispectral observations of the hidden scene and a camera to capture the indirect scattered light. Our results show that BSS was successful in suppressing even overwhelming amounts of clutter from simple scenes. Performance deteriorates slightly with clutter intensity, but scales well with the number of different spectra measurements. This project was successful in determining the contributions and limitations of BSS in multispectral NLoS imaging, paving the way for future advancements in this area.