Single-shot inline holography using a physics-aware diffusion model.

Area of Science:

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

Background:

• Inline holography offers compact and portable imaging solutions ideal for field applications.
• Holographic reconstruction from single-shot measurements presents significant challenges, particularly in reducing twin-image noise.
• Existing methods often require extensive training data or complex parameter tuning.

Purpose of the Study:

• To introduce a novel unsupervised algorithm, the physics-aware diffusion model for digital holographic reconstruction (PadDH), for improved single-shot holographic imaging.
• To demonstrate PadDH's ability to overcome the limitations of traditional holographic reconstruction techniques.
• To validate the algorithm's effectiveness in reducing twin-image contamination and enhancing reconstruction quality.

Main Methods:

• Development of PadDH, an unsupervised algorithm integrating physical principles with a pre-trained diffusion model.
• Implementation of PadDH for digital holographic reconstruction without requiring a specific holographic training dataset.
• Experimental validation using both synthetic and real-world holographic data.

Main Results:

• PadDH effectively reduces twin-image contamination in holographic reconstructions.
• The algorithm generates high-quality holographic images from single-shot measurements.
• PadDH demonstrates significant parameter reduction compared to other methods.

Conclusions:

• PadDH represents a significant advancement in unsupervised holographic imaging, leveraging pre-trained diffusion models.
• The algorithm's physics-aware approach enhances reconstruction quality and reduces artifacts.
• PadDH offers a robust and efficient solution for on-site and field holographic applications.