Clutter rejection in passive non-line-of-sight imaging via blind multispectral unmixing

Area of Science:

• Optics and Photonics
• Computer Vision
• Signal Processing

Background:

• Passive non-line-of-sight (NLOS) imaging uses scattered light to see around obstacles.
• Unwanted light sources (clutter) often overpower the desired imaging signal.
• Spectral content analysis has shown promise in mitigating clutter.

Purpose of the Study:

• To quantify the effectiveness of preconditioning and unmixing algorithms for blind source separation (BSS) in passive multispectral NLOS imaging.
• To evaluate the impact of these algorithms on image reconstruction quality under varying scene conditions.

Main Methods:

• Utilized an OLED television as a controlled source for both desired signals and clutter.
• Employed BSS methods with various preconditioning and unmixing algorithms.
• Conducted controlled experiments to test algorithm performance across different scene conditions.

Main Results:

• Preconditioning significantly reduces clutter power and correlation, proving vital for signal isolation.
• The choice of unmixing algorithm critically influences the quality of the reconstructed image.
• Effective image retrieval was achieved with clutter signals up to 670 times stronger than the desired image.

Conclusions:

• Optimizing preconditioning and unmixing algorithms is crucial for successful passive multispectral NLOS imaging.
• The developed methods demonstrate robust performance even in the presence of substantial optical clutter.
• This research advances the capabilities of NLOS imaging for practical applications.