Self-learning based Fourier ptychographic microscopy.

Yongbing Zhang,Qionghai Dai,Laura Waller,Weixin Jiang,Lei Tian

doi:10.1364/oe.23.018471

Abstract

Fourier Ptychographic Microscopy (FPM) is a newly proposed computational imaging method aimed at reconstructing a high-resolution wide-field image from a sequence of low-resolution images. These low-resolution images are captured under varied illumination angles and the FPM recovery routine then stitches them together in the Fourier domain iteratively. Although FPM has achieved success with static sample reconstructions, the long acquisition time inhibits real-time application. To address this problem, we propose here a self-learning based FPM which accelerates the acquisition and reconstruction procedure. We first capture a single image under normally incident illumination, and then use it to simulate the corresponding low-resolution images under other illumination angles. The simulation is based on the relationship between the illumination angles and the shift of the sample's spectrum. We analyze the importance of the simulated low-resolution images in order to devise a selection scheme which only collects the ones with higher importance. The measurements are then captured with the selection scheme and employed to perform the FPM reconstruction. Since only measurements of high importance are captured, the time requirements of data collection as well as image reconstruction can be greatly reduced. We validate the effectiveness of the proposed method with simulation and experimental results showing that the reduction ratio of data size requirements can reach over 70%, without sacrificing image reconstruction quality.

Highlights

The development of traditional optical systems is constrained by space-bandwidth product (SBP) [1], which forces the user to choose between high-resolution or large field-of-view (FOV)
We validate our method in theory and practice as compared to traditional Fourier Ptychographic Microscopy (FPM), reducing the time needed for data collection as well as image reconstruction
The time reduction of the images in the data-collection procedure is more than 70% for the final scan, since the discarded images, which are mostly captured under the illumination of marginal LEDs, require more exposure time than the central brightfield LEDs

Summary

Introduction

The development of traditional optical systems is constrained by space-bandwidth product (SBP) [1], which forces the user to choose between high-resolution or large field-of-view (FOV). Large SBP images generally cannot be captured with a single shot, they can be reconstructed from a sequence of low-resolution images by computational methods. Multiplexing can be used for the RGB spectral channels, as shown by Dong et al [12] These variations on FPM have found applications in quantitative phase imaging [13], gigapixel microscopy [14], high-resolution fluorescence imaging [15] and more [16,17,18]. In this paper, considering the relationship between a high-resolution image spectrum and its corresponding low-resolution spectra [19,20,21,22], we add a decision-making procedure to the FPM routine, which helps to select the LEDs corresponding to the most informative measurements in order to only capture the important information.

Fourier ptychographic microscopy

Adaptive Fourier ptychography

Principle of self-learning based FPM

Simulation results

Experimental results

Conclusion and discussion