Single-pixel deep phase-shifting incoherent digital holography

Area of Science:

• Optics and Photonics
• Digital Holography
• Computational Imaging

Background:

• Incoherent digital holography broadens applications by relaxing light source coherence requirements.
• Traditional methods often require complex setups and significant data acquisition.
• Area array sensors can be a limitation in certain digital holography applications.

Purpose of the Study:

• To develop a novel phase-shifting network for incoherent digital holography.
• To enable high-fidelity 3D reconstruction using minimal holographic data.
• To improve object light energy utilization and circumvent sensor constraints.

Main Methods:

• Designed a Multi-head attention single-pixel (MHASP) phase-shifting network.
• Network predicts three phase-shifted interferograms (0, 2/3π, 4/3π) from 1D input.
• Employed back propagation algorithm for holographic reconstruction, eliminating DC and twin terms.
• Calibrated back propagation distance for 3D information preservation.

Main Results:

• Achieved high-fidelity holographic reconstruction with minimal data volume.
• MHASP network effectively replaced physical phase-shifting operations.
• Demonstrated preservation of 3D information through calibrated back propagation.
• Improved utilization of object light energy compared to conventional methods.

Conclusions:

• The MHASP phase-shifting network significantly advances incoherent digital holography.
• This method allows for high-precision 3D imaging with reduced data requirements.
• The approach expands the applicability of digital holography, particularly in 3D imaging scenarios.