Point Spread Function Engineering Using Implicit Neural Representations

Abstract

Point spread function (PSF) engineering through pupil plane modulation is a technique used in microscopy to achieve specific imaging properties, such as depth encoding or extended depth of field. Existing PSF design methods often rely on extensive domain knowledge and task-specific basis functions, making it difficult to generalize across different applications. We treat the PSF engineering task as a phase retrieval problem and propose a neural field pupil design method that optimizes a phase profile for any arbitrary, user-defined 3D PSF distribution. This provides a flexible framework for 3D PSF engineering for various applications with implicit regularization that proves robust to initialization compared to pixel-wise optimization methods