Self-learning based Fourier ptychographic microscopy.

Area of Science:

• Computational imaging
• Optical microscopy
• Image reconstruction

Background:

• Fourier Ptychographic Microscopy (FPM) reconstructs high-resolution images from low-resolution data.
• Current FPM methods face limitations due to long acquisition times, hindering real-time applications.
• Static sample reconstructions are successful, but dynamic imaging remains a challenge.

Purpose of the Study:

• To accelerate Fourier Ptychographic Microscopy (FPM) acquisition and reconstruction using a self-learning strategy.
• To enable real-time or near real-time high-resolution imaging applications.
• To reduce data acquisition and processing time without sacrificing image quality.

Main Methods:

• Developed a self-learning FPM method utilizing a single normally incident illumination image.
• Simulated low-resolution images under varied illumination angles based on spectral shift principles.
• Implemented an importance-based selection scheme to capture only critical measurements.
• Performed FPM reconstruction using the selected, high-importance measurements.

Main Results:

• Achieved significant acceleration in both data acquisition and FPM reconstruction.
• Demonstrated data size reduction exceeding 70% through the importance-based selection scheme.
• Validated the method's effectiveness via simulation and experimental results.
• Maintained high image reconstruction quality comparable to traditional FPM.

Conclusions:

• The proposed self-learning FPM effectively accelerates imaging processes.
• Intelligent data selection is key to reducing acquisition time and computational load.
• This method broadens the potential for real-time FPM applications in microscopy.