Adaptive coded phase mask design and high-quality image reconstruction for interference-less coded aperture...

Area of Science:

• Optics and Photonics
• Computational Imaging
• Machine Learning Applications

Background:

• Coded aperture correlation holography offers non-scanning, incoherent 3D imaging.
• Challenges include system noise encoded by the phase mask and background artifacts from reconstruction.

Purpose of the Study:

• To develop a deep learning-based method for simultaneous mitigation of system noise and background components in coded aperture correlation holography.
• To enhance the signal-to-noise ratio (SNR) of reconstructed 3D images.

Main Methods:

• A two-sub-network deep learning approach: a coded phase mask design sub-network and an image reconstruction sub-network.
• The design sub-network generates an adaptive coded phase mask using object frequency distribution.
• The reconstruction sub-network maps hologram autocorrelations to the object, incorporating physical knowledge to suppress background.

Main Results:

• The proposed method effectively suppresses system noise and background components.
• Significant improvement in the signal-to-noise ratio (SNR) of reconstructed images was demonstrated.
• The deep learning approach enhances adaptability and interpretability of the holographic reconstruction process.

Conclusions:

• The deep learning-based method successfully addresses key challenges in coded aperture correlation holography.
• This technique offers a robust solution for noise and background reduction, leading to higher quality 3D imaging.
• The study highlights the potential of AI in advancing holographic imaging techniques.