Optical-flow-based Hybrid Neural Operator for Unsupervised Flow Estimation in Particle Image Velocimetry

Particle image velocimetry is a widely-used experimental technique for measuring fluid flow, but accurate flow field estimation from particle images remains challenging. Traditional methods, such as cross-correlation and optical flow, are either resolution-limited or computationally expensive, and deep learning approaches often suffer from discretization errors, spectral bias, or a heavy reliance on labeled data. In this study, we propose the first framework that integrates operator learning wit