Learning-based correction with Gaussian constraints for ghost imaging through dynamic scattering media.

Area of Science:

• Optics and Photonics
• Machine Learning
• Image Reconstruction

Background:

• Ghost imaging (GI) traditionally struggles with dynamic scattering media, limiting its practical applications.
• Existing methods for correcting scattering effects in GI are often complex and data-intensive.

Purpose of the Study:

• To develop a novel learning-based correction method for ghost imaging through dynamic scattering media.
• To improve the quality and efficiency of image recovery in challenging optical environments.

Main Methods:

• Utilized deep neural networks with Gaussian constraints to learn scattering mechanisms.
• Developed a correction method to rectify dynamic scaling factors in the optical channel.
• Ensured corrected realizations follow a Gaussian distribution for high-quality image recovery.

Main Results:

• Successfully recovered high-quality ghost images through dynamic scattering media.
• Demonstrated the effectiveness and robustness of the proposed learning-based correction method.
• Achieved image recovery using half the number of realizations compared to traditional temporally corrected GI methods.

Conclusions:

• The proposed learning-based method offers a novel insight into ghost imaging.
• This approach is a promising tool for optical imaging through dynamic scattering media.
• The method enhances image quality and reduces data requirements in challenging imaging scenarios.