Abstract
Fourier ptychographic microscopy is a technique that achieves a high space-bandwidth product, i.e. high resolution and high field-of-view. In Fourier ptychographic microscopy, variable illumination patterns are used to collect multiple low-resolution images. These low-resolution images are then computationally combined to create an image with resolution exceeding that of any single image from the microscope. Due to the necessity of acquiring multiple low-resolution images, Fourier ptychographic microscopy has poor temporal resolution. Our aim is to improve temporal resolution in Fourier ptychographic microscopy, achieving single-shot imaging without sacrificing space-bandwidth product. We use example-based super-resolution to achieve this goal by trading off generality of the imaging approach. In example-based super-resolution, the function relating low-resolution images to their high-resolution counterparts is learned from a given dataset. We take the additional step of modifying the imaging hardware in order to collect more informative low-resolution images to enable better high-resolution image reconstruction. We show that this "physical preprocessing" allows for improved image reconstruction with deep learning in Fourier ptychographic microscopy. In this work, we use deep learning to jointly optimize a single illumination pattern and the parameters of a post-processing reconstruction algorithm for a given sample type. We show that our joint optimization yields improved image reconstruction as compared with sole optimization of the post-processing reconstruction algorithm, establishing the importance of physical preprocessing in example-based super-resolution.
Highlights
IntroductionA subset of machine learning, has shown remarkable results in the interpretation of data
Deep learning, a subset of machine learning, has shown remarkable results in the interpretation of data
We assume that the light-emitting diode (LED) is sufficiently far away that the light entering the sample can be approximated as a plane wave
Summary
A subset of machine learning, has shown remarkable results in the interpretation of data. The “super-resolution" problem of converting low-resolution images to higher resolution has been attempted with machine learning [4,5,6,7,8,9,10,11,12]. We assume that the LED is sufficiently far away that the light entering the sample can be approximated as a plane wave. The spatial profile of the plane wave from the LED entering the sample is described by ei2πuìl ·rì where uìl · x = ul,x is the spatial frequency in the x-direction and uìl · y = ul,y is the spatial frequency in the y-direction.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have