End-to-end computational ghost imaging method that suppresses atmospheric turbulence

Area of Science:

• Optics
• Computational Imaging
• Machine Learning

Background:

• Atmospheric turbulence significantly degrades image quality.
• Traditional ghost imaging suppresses turbulence but lacks stability and high-quality reconstruction under extreme conditions.
• Computational ghost imaging (CGI) offers a potential solution but requires further optimization.

Purpose of the Study:

• To investigate the effectiveness of computational ghost imaging combined with neural networks for atmospheric turbulence suppression.
• To develop an end-to-end neural network approach for direct image reconstruction from bucket signals.
• To analyze factors affecting imaging and demonstrate robust performance under challenging conditions.

Main Methods:

• Simulating atmospheric turbulence using phase screens.
• Employing computational ghost imaging principles for image acquisition simulation.
• Developing and applying an end-to-end neural network to process bucket signals for direct image reconstruction.
• Analyzing performance at low sampling rates and under simulated extreme conditions.

Main Results:

• The end-to-end neural network successfully reconstructs target images directly from processed bucket signals.
• The proposed method eliminates the need for traditional correlation calculations, significantly saving reconstruction time.
• Good image reconstruction results were achieved even at low sampling rates and under simulated extreme atmospheric turbulence conditions.
• The neural network approach demonstrates improved stability and quality compared to conventional ghost imaging.

Conclusions:

• End-to-end neural networks offer a powerful and efficient method for enhancing ghost imaging under atmospheric turbulence.
• This approach overcomes limitations of traditional methods, enabling high-quality image reconstruction in challenging environments.
• The findings suggest a promising direction for developing robust imaging systems for applications affected by atmospheric disturbances.