Computational ghost imaging for atmospheric turbulence using model-driven and data-driven deep learning.

Area of Science:

• Optics and Photonics
• Computational Imaging
• Machine Learning Applications

Background:

• Atmospheric turbulence severely distorts images, challenging conventional imaging techniques.
• Computational ghost imaging (CGI) offers turbulence resistance but is limited by sampling rates.
• Existing deep learning methods for CGI lack generalizability and interpretability.

Purpose of the Study:

• To develop a novel computational ghost imaging method for atmospheric turbulence.
• To enhance image reconstruction performance under low-sampling conditions.
• To combine model-driven and data-driven deep learning for improved generalizability and interpretability.

Main Methods:

• Integration of model-driven and data-driven deep learning strategies for CGI.
• Leveraging implicit features from data-driven methods and generalization from model-driven approaches.
• Utilizing second-order correlation algorithms for object reconstruction.

Main Results:

• The proposed hybrid deep learning CGI method demonstrates robustness across various sampling ratios.
• The approach effectively mitigates image distortions caused by atmospheric turbulence.
• Simulation and experimental results confirm the method's high-quality imaging capabilities.

Conclusions:

• The integrated model- and data-driven deep learning approach offers a superior solution for CGI in turbulent environments.
• This method overcomes the limitations of conventional CGI and pure data-driven techniques.
• The findings present an effective pathway for achieving high-fidelity imaging under atmospheric turbulence.