Adaptive denoising methods in Fourier ptychographic microscopy

Abstract

Fourier ptychographic microscopy (FPM) is a computational imaging technique that addresses several issues in optical systems such as low space-bandwidth product (SBP) and high-cost materials. Here, we compared three different algorithms: the traditional FPM algorithm, an adaptive denoising FPM algorithm that introduced a noise discrimination matrix, and an FPM algorithm that used thresholding methods. Gaussian noise of different standard deviation was introduced and the peak signal-to-noise ratio (PSNR) values of the recovered intensity and phase were calculated as a metric for comparison. All simulations were conducted and tested in MATLAB. It was observed that the traditional FPM algorithm recovered the intensity with the highest PSNR and is thus the most robust in terms of recovered intensity. On the other hand, the thresholding method yielded the recovered phase with the highest PSNR, making it the most robust in terms of phase recovery. The adaptive denoising with noise discrimination factor was evaluated to perform in between the two aforementioned algorithms.