Single-shot Fresnel incoherent correlation holography via deep learning based phase-shifting technology

Area of Science:

• Optics and Photonics
• Computational Imaging
• Machine Learning Applications

Background:

• Fresnel incoherent correlation holography (FINCH) enables non-scanning 3D imaging but requires complex phase-shifting for accurate reconstruction.
• Existing FINCH methods face limitations in real-time performance and experimental simplicity due to phase-shifting requirements.

Purpose of the Study:

• To develop a rapid and high-precision single-shot FINCH method using deep learning for phase-shifting.
• To reduce experimental complexity and improve the real-time capabilities of FINCH.

Main Methods:

• A deep learning-based phase-shifting network (FINCH/DLPS) was designed to predict phase-shifted interferograms from a single input.
• Conventional three-step phase-shifting algorithms were applied to the predicted interferograms.
• Back propagation algorithms were used for high-precision image reconstruction.

Main Results:

• The FINCH/DLPS method successfully reconstructed high-precision 3D images from a single interferogram.
• The method effectively removed DC and twin terms, simplifying the experimental setup.
• Feasibility was validated using the MNIST dataset, demonstrating retention of 3D information.

Conclusions:

• The proposed FINCH/DLPS method offers a superior alternative for rapid and precise 3D holographic imaging.
• Deep learning-based phase-shifting significantly reduces experimental complexity and enhances real-time performance.
• This approach holds promise for advanced applications in 3D imaging and microscopy.