ELFPIE: an error-laxity Fourier ptychographic iterative engine

• DOI: 10.1016/j.sigpro.2023.109088
• Publication Date: 2023-04-29T01:09:54Z
• Authors: Shuhe Zhang; Tos T.J.M. Berend...; Jinhua Zhou
• Keywords: Gaussian Noise; Noise power

We present a simple but efficient and robust reconstruction algorithm for Fourier ptychographic microscopy, termed error-laxity iterative engine (Elfpie), that is simultaneously to (1) noise signal (including Gaussian, Poisson, salt & pepper noises), (2) problematic background illumination problem, (3) vignetting effects (4) misaligning of LED positions, without the need calibrating or recovering these system errors. In Elfpie, we embed inverse problem FPM under framework feature extraction/recovering propose new image gradient-based data fidelity cost function regularized by global second-order total-variation regularization. The closed-form complex gradient derived back-propagated using AdaBelief optimizer with an adaptive learning rate. Elfpie was tested on both simulation experimental data. general, compared against SOTA methods, Gaussian 100 times larger noise, 1000 Poisson 10 noise. able reconstruct high-fidelity samples position misalignments up 2 mm. It can also bypass effect, which all methods fail sample pattern. MATLAB code ELFPIE available Github.