SNR-based adaptive acquisition method for fast Fourier ptychographic microscopy

TL;DR

This paper introduces an SNR-based adaptive acquisition method for Fourier ptychographic microscopy that enhances efficiency by automatically selecting optimal raw images, significantly reducing data collection time without increasing computational complexity.

Contribution

It proposes a novel SNR-based adaptive collection strategy that improves FPM efficiency and achieves over 90% reduction in image acquisition compared to traditional methods.

Findings

Achieved over 90% reduction in the number of collected images.

Demonstrated effectiveness on USAF targets and biological samples.

Compatible with existing EPRY-FPM algorithm without added complexity.

Abstract

Fourier ptychographic microscopy (FPM) is a computational imaging technique with both high resolution and large field-of-view. However, the effective numerical aperture (NA) achievable with a typical LED panel is ambiguous and usually relies on the repeated tests of different illumination NAs. The imaging quality of each raw image usually depends on the visual assessments, which is subjective and inaccurate especially for those dark field images. Moreover, the acquisition process is really time-consuming.In this paper, we propose a SNR-based adaptive acquisition method for quantitative evaluation and adaptive collection of each raw image according to the signal-to-noise ration (SNR) value, to improve the FPM's acquisition efficiency and automatically obtain the maximum achievable NA, reducing the time of collection, storage and subsequent calculation. The widely used EPRY-FPM algorithm is applied without adding any algorithm complexity and computational burden. The performance has been demonstrated in both USAF targets and biological samples with different imaging sensors respectively, which have either Poisson or Gaussian noises model. Further combined with the sparse LEDs strategy, the number of collection images can be shorten to around 25 frames while the former needs 361 images, the reduction ratio can reach over 90%. This method will make FPM more practical and automatic, and can also be used in different configurations of FPM.