Uncertainty-aware physics-informed high-fidelity ghost imaging in ambient illumination.

Area of Science:

• Optics and Photonics
• Computational Imaging
• Physics-Informed Machine Learning

Background:

• Ghost imaging offers high sensitivity but is vulnerable to ambient light, which degrades image quality.
• Existing methods struggle to maintain reconstruction fidelity under environmental perturbations.

Purpose of the Study:

• To develop an uncertainty-aware, physics-informed framework for robust ghost imaging.
• To address distortions caused by ambient light and improve image reconstruction fidelity.

Main Methods:

• Implemented a dual-branch neural network architecture.
• Employed a progressive training strategy to disentangle image reconstruction from noise suppression.
• Integrated physics-informed constraints to characterize target and mask distortions.

Main Results:

• Achieved high-fidelity ghost imaging even with significant ambient light interference.
• Demonstrated reduced reliance on precise measurement matrices.
• Enhanced stability and fidelity of image reconstruction compared to conventional methods.

Conclusions:

• The proposed framework effectively mitigates ambient light perturbations in ghost imaging.
• This approach enables reliable ghost imaging in complex and noisy environments.
• Offers a pathway for more robust and adaptable ghost imaging systems.