Hyperuniform illumination subsampling method for Fourier ptychographic microscopy

Abstract

Fourier ptychographic microscopy (FPM) is a novel computational microscopy technique, where Angular-varying LED illumination and phase retrieval techniques are used to stitch a wide field-of-view (FOV) and high-resolution (HR) image. However, the imaging process of FPM needs long acquisition and reconstruction time, which limits the application of FPM in dynamic imaging. Here, we propose a hyperuniform illumination subsampling method to generate 2D hyperuniform LED illumination patterns with grid and collision distance constraints, termed HUD-FPM, to improve FPM imaging efficiency. The corresponding LED element can be lighted up in sequence, and the low-resolution (LR) images can be captured and reconstructed. We verify the effectiveness of the proposed method through simulations and experiments. The proposed method can achieve reconstruction quality close to tranditional sampling FPM while reducing the number of image captures by about 67.2%. Images reconstructed through hyperuniform illumination mode have lower root mean square error and exhibit finer details than commonly used sparse sampling methods at high frequencies. Further, the same conclusion can be also applied to the non-uniform and high-NA quasi-dome LED illuminator design.