End-to-End Inverse Designed Single-Layered Metasurface for High-Compression Snapshot RGB-Achromatic Full-Stokes Polarization Imaging

Abstract

Snapshot full-Stokes polarimetry across multiple wavelengths remains challenging because conventional architectures rely on multiplexed measurements and bulky optics. We present an end-to-end inverse designed single-layered metasurface that reconstructs RGB full-Stokes images from a snapshot sensor measurement. A metasurface modeled by a multilayer perceptron (MLP) is employed to encode the full-Stokes polarization information. The system jointly optimizes a differentiable metasurface frontend with a U-Net backend. On a real-world dataset, our design achieves a high compression ratio of 12 for snapshot RGB-achromatic polarization imaging using a single-layered metasurface. These results show that end-to-end optical-digital co-design enables high-compression snapshot full-Stokes polarization imaging with a compact footprint.