Efficient color imaging through unknown opaque scattering layers via physics-aware learning.

Area of Science:

• Optics and Photonics
• Artificial Intelligence in Imaging
• Computational Imaging

Background:

• Color imaging through scattering media is a significant challenge in optical imaging.
• Deep learning (DL) approaches have enhanced optical imaging capabilities by integrating physics theories.
• Reconstructing complex objects obscured by opaque scattering layers requires advanced techniques.

Purpose of the Study:

• To propose an efficient color imaging method for reconstructing objects behind unknown opaque scattering layers.
• To achieve high reconstruction fidelity in spatial structure and accurate color restoration.
• To leverage scattering redundancy and physics-aware DL for robust generalization.

Main Methods:

• Development of a physics-aware deep learning approach for optical imaging.
• Training the model with a single diffuser to reconstruct complex objects.
• Utilizing scattering redundancy to extract more information from scattered light.

Main Results:

• High reconstruction fidelity in spatial structure of complex objects.
• Accurate restoration of color information from scattered light.
• Robust generalization capability for reconstructing objects behind unknown scattering layers.

Conclusions:

• The proposed method enables efficient color imaging through dynamic scattering media.
• Physics-aware DL offers a powerful approach for solving complex inverse problems in imaging.
• This work provides a valuable reference for advanced optical imaging applications.