Plug-and-play pixel super-resolution phase retrieval for digital holography.

Area of Science:

• Optical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Current optical imaging detectors often trade resolution for signal-to-noise ratio by increasing pixel size.
• This limitation hinders advancements in high-throughput holographic imaging.
• Existing pixel super-resolution (PSR) techniques face a trade-off between reconstruction fidelity and the degree of super-resolution.

Purpose of the Study:

• To introduce a high-fidelity pixel super-resolution (PSR) phase retrieval method using plug-and-play optimization (PNP-PSR).
• To address the limitations of existing PSR methods by improving reconstruction quality and super-resolution ratio.
• To enhance signal-to-noise ratio and resolution in optical imaging.

Main Methods:

• Developed PNP-PSR, a method decomposing PSR reconstruction into sub-problems using a generalized alternating projection framework.
• Employed an alternating projection operator for measurement fidelity and an enhancing neural network for statistical prior regularization.
• Integrated advantages of individual operators for efficiency and noise robustness.

Main Results:

• PNP-PSR achieves high-fidelity reconstruction in pixel super-resolution phase retrieval.
• The method demonstrates superior performance in both resolution enhancement and noise suppression compared to existing techniques.
• Extensive experiments validate the efficiency and noise robustness of PNP-PSR.

Conclusions:

• PNP-PSR offers a significant advancement in holographic imaging and optical microscopy.
• The plug-and-play optimization approach effectively balances reconstruction quality and super-resolution.
• PNP-PSR provides a robust solution for high-throughput imaging applications requiring high resolution and low noise.