Linear-space-variant model for Fourier ptychographic microscopy.

Area of Science:

• Optics and Photonics
• Microscopy Techniques
• Computational Imaging

Background:

• Fourier ptychographic microscopy (FPM) reconstructions often suffer from vignetting artifacts, leading to uneven color blocks and stitching issues.
• Current FPM methods rely on image segmentation and data discarding to mitigate these artifacts, impacting reconstruction quality.
• The underlying cause is the mismatch between the acquired data and the linear-space-invariant (LSI) forward model, neglecting quadratic phase modulation.

Purpose of the Study:

• To address the persistent vignetting artifacts in FPM.
• To develop a novel FPM model that accurately describes the data acquisition process.
• To improve the quality and reliability of FPM reconstructions without manual data curation.

Main Methods:

• Re-derivation of a linear-space-variant (LSV) forward model for FPM.
• Development of a corresponding loss function, termed LSV-FPM.
• Optimization using the LSV-FPM model to process FPM data.

Main Results:

• LSV-FPM effectively removes wrinkle artifacts caused by vignetting.
• The new method eliminates the need for image segmentation or discarding problematic data.
• Successful validation of LSV-FPM on data from both 4f and finite conjugate single-lens systems.

Conclusions:

• The proposed LSV-FPM model accurately accounts for quadratic phase modulation during FPM data acquisition.
• LSV-FPM offers a robust solution for artifact removal, significantly enhancing FPM reconstruction quality.
• This advancement provides a more satisfying user experience and reliable imaging results for FPM researchers.