Computational Microscopy Via Fourier Ptychography

Fourier ptychographic microscopy(FPM)is a recently developed computational imaging technique that combines phase retrieval and synthetic aperture to overcome the inherent trade-offs between the large field-of-view(FOV)and high-resolution(HR)in conventional microscope,achieving gigapixels image with the improved spatial bandwidth product(SBP)for dozens of times.In addition,FPM offeres a new method for quantitative phase imaging and digital aberation removal,which has many advantages such as low-cost,label-free and high sensitivity.In recent years,FPM has been found many applications in quantitative phase microscopy,live-cell imaging and super-resolution microscopy.In this paper,main advantages of the FPM approach in SBP improvement and phase retrieval are fully discussed,and some critical factors that may affect FPM's performance are analyzed in-depth.The main contents of the dissertation are summarized as follows:1.A LED-based FPM experimental setup is designed,achieving a spatial resolution of 1.55?m with an illumination wavelenghth of 631 nm and a 4×/0.1NA objective lens.This system improves the SBP by sixteen-fold,overcomeing the trade-offs between the large FOV and HR in conventional microscope.The potential of resolution enhancement,quantitative phase and aberration removal of the FPM approach is demonstrated by the experimental results of biological samples.2.An adaptive acquisition method based on peak signal-to-noise ratio(PSNR)is proposed to improve the FPM's data acquisition speed.The FPM's original dataset is successfully compressed from 225 raw images to 65 images within the permissible resolution loss.Besides,the image degradation caused by insufficient aperture overlap,CCD sampling and sample thickness is analyzed in-depth.Finally,the sub-pixel sampling method is proposed to address phase artifacts and spectrum distortion caused by undersampling in the reconstruction process.3.The image degradation caused by LED position misalignment,intensity fluctuation and background noise is fully discussed.A data pre-processing method based on the least square is proposed,achieving noise suppression at full frame in the captured raw images.In addition,a set of binary pixel masks are employed to eliminate the negetive effects of stray light on the dark-field raw images,greatly improving the consistency between the captured raw images.4.A PIE-based system calibration algorithm,termed SC-FPM,is proposed to address LED position misalignment and intensity fluctuation.Through a combination of the simulated annealing and non-linear programming,SC-FPM enables to correct frequency positions of each LED element.,and to suppress algorithm oscillation near the convergency value by adaptive step-size strategy,significantly improving the noise-robust capability of the FPM's reconstruction algorithm..5.A high resolution FPM experimental setup(HR-FPM),equipped with a hemispherical LED condenser that offers 0.95 numerical aperture and 465 nm illumination wavelength,is designed to address the limited synthetic aperture in the planar LED-based FPM system.HR-FPM obtains 1.05 synthetic aperture and a sub-wavelength resolution of 488 nm with a 4×/0.1NA objective lens,achieving gigapixels SBP in the experiment,demonstrating that HR-FPM successfully combines the joint advantages of large FOV in low-NA objective and high-resolution in high-NA objective.