Implicitly Learned Neural Phase Functions for Basis-Free Point Spread Function Engineering

📅 2024-10-07
🏛️ arXiv.org
📈 Citations: 0
Influential: 0
📄 PDF
🤖 AI Summary
In computational imaging, designing optical phase masks to achieve target point spread functions (PSFs) faces fundamental challenges—including mathematical ill-posedness, strong dependence on handcrafted physical basis functions, and poor generalization across diverse PSF specifications. Method: We propose an end-to-end phase design framework based on implicit neural representations (INRs), directly parameterizing the continuous phase field as a coordinate-to-phase neural network—bypassing explicit basis function constraints—and integrating differentiable Fourier-domain optimization to ensure optical feasibility and high-fidelity PSF synthesis. Results: Experiments demonstrate that our method significantly outperforms pixel-wise optimization in reconstruction quality and cross-task generalization for multi-focus, annular, and asymmetric PSF targets. It enables flexible, high-precision PSF engineering without prior structural assumptions, establishing a new paradigm for neural imaging and bio-photonics applications such as fluorescence microscopy.

Technology Category

Application Category

📝 Abstract
Point spread function (PSF) engineering is vital for precisely controlling the focus of light in computational imaging, with applications in neural imaging, fluorescence microscopy, and biophotonics. The PSF is derived from the magnitude of the Fourier transform of a phase function, making the construction of the phase function given the PSF (PSF engineering) an ill-posed inverse problem. Traditional PSF engineering methods rely on physical basis functions, limiting their ability to generalize across the range of PSFs required for imaging tasks. We introduce a novel approach leveraging implicit neural representations that significantly outperforms existing pixel-wise optimization methods in phase function quality.
Problem

Research questions and friction points this paper is trying to address.

Light Diffraction Patterns
Computational Imaging
Biophotonics
Innovation

Methods, ideas, or system contributions that make the work stand out.

Neural Networks
Phase Pattern Learning
Image Diffusion Optimization
🔎 Similar Papers
No similar papers found.
A
Aleksey Valouev
Boston University