Adaptive windowing for photon-efficient non-line-of-sight imaging under high ambient light

Area of Science:

• Optics and Photonics
• Computational Imaging
• Computer Vision

Background:

• Non-line-of-sight (NLOS) imaging reconstructs hidden object shapes but is limited by weak signals in ambient light.
• Low signal-to-background ratio (SBR) severely degrades practical NLOS imaging performance.

Purpose of the Study:

• To develop a robust NLOS imaging method for high-quality reconstruction under extreme low SBR and photon-starved conditions.
• To improve imaging performance by effectively suppressing background noise and amplifying weak signals.

Main Methods:

• Introduced adaptive windowing non-line-of-sight (AW-NLOS) imaging.
• Utilized probabilistic models for signal/noise detection and spatio-temporal clustering for signal amplification.
• Employed matched filtering for window size estimation and per-pixel adaptive short-duration range windows.
• Integrated total variation (TV) regularization for enhanced reconstruction.

Main Results:

• Achieved reliable reconstruction at SBR as low as 2.12 and signal flux of 0.02 photons per pixel.
• Demonstrated significant improvement in structural similarity index (SSIM) by approximately 0.5 compared to conventional methods.
• Showcased high-quality reconstruction under extreme background noise and limited photon detection efficiency.

Conclusions:

• AW-NLOS imaging provides a novel paradigm for precise reconstruction in challenging photon-starved and high-noise NLOS scenarios.
• The method enables robust and high-fidelity imaging where conventional techniques fail.