Ghost imaging based on Y-net: a dynamic coding and decoding approach.

Area of Science:

• Optics and Photonics
• Machine Learning
• Image Reconstruction

Background:

• Ghost imaging typically requires deterministic illumination and multiple exposures.
• Deep learning has shown promise in enhancing optical imaging techniques.
• Existing methods face limitations in flexibility and efficiency.

Purpose of the Study:

• To develop a novel dynamic decoding deep learning framework for ghost imaging.
• To enable ghost imaging under both deterministic and indeterministic illumination conditions.
• To improve image quality and reduce radiation damage in high-resolution imaging.

Main Methods:

• A novel dynamic decoding deep learning framework (Y-net) was developed.
• The network performs end-to-end image reconstruction directly from detector data.
• The method accommodates variations between experimental and simulated speckle patterns.

Main Results:

• The proposed Y-net framework successfully performs ghost imaging with single illumination.
• The scheme is effective under both deterministic and indeterministic illumination.
• Reconstructed images show improved quality, with potential for reduced radiation dose.

Conclusions:

• The developed deep learning framework offers a flexible and efficient ghost imaging solution.
• This approach has significant implications for high-resolution X-ray imaging applications.
• The method paves the way for advanced imaging techniques with reduced experimental constraints.