Research map / 3D vision / Depth and stereo
Event Ellipsometer: Event-based Mueller-Matrix Video Imaging
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Event Ellipsometer captures normalized Mueller-matrix video at 30 fps by rotating QWPs in front of an LED source and an event camera, then reconstructing per-pixel Mueller matrices from event time differences with spatio-temporal propagation.
Problem
Conventional ellipsometers need many rotation angles and long acquisition, so Mueller-matrix imaging has been limited to static scenes; single-shot alternatives are restricted to planar scenes, sacrifice resolution, or need custom optics.
Approach
Fast-rotating QWPs on the light source and camera sides modulate polarization so that an event camera's asynchronous intensity-change events encode the Mueller matrix via an ellipsometric-event image formation model; reconstruction uses weighted least squares with Cloude filtering and PatchMatch-style spatio-temporal propagation.
Key results
- Mueller-matrix video reconstruction at 30 fps with a mean-squared error of 0.045 on materials with known Mueller matrices.
- On synthetic data, SVD initialization yields reconstruction error 0.11 while the full pipeline reaches 0.04.
- Reconstruction of 500x500, 30-frame event sequences takes 62 s per-pixel plus 68 s propagation on a Ryzen 7 7800X3D CPU.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use the full reconstruction pipeline instead of SVD initialization alone (ablation) | Event Ellipsometer · Synthetic scene | reconstruction error: 0.11 -> 0.04 | Helped |
| use an event camera with rotating QWPs to capture Mueller-matrix video | Event Ellipsometer · Dynamic scenes | frame rate: 30 fps | Helped |
| use event-based capture instead of conventional frame-based ellipsometry | Event Ellipsometer · Known optical elements | capture duration: 33ms; several minutes | Helped |
Limitations
- Reconstructs only normalized Mueller matrix M/M(0,0), not full Mueller matrix, since event cameras detect intensity changes only.
- CPU-only reconstruction makes the pipeline non-real-time.
- Event bandwidth and contrast threshold trade-off limits detection of weakly polarized signals.
Experimental setup
Datasets: Mitsuba3 synthetic renders with real-world polarimetric BRDFs (blue silicone, brass)
Benchmarks: synthetic rendered event sequences known optical elements (air, LP, QWP) real-world scenes (metal plate, gelatine disk, tape, face, hair, HDR scene)
Metrics: mean squared error reconstruction error (mean absolute error)
Baselines: single-shot Mueller matrix imaging (Yu et al. [49]) conventional rotating-optics ellipsometer
Builds on: PatchMatchStereo Cloude Mueller matrix filtering DVS-Voltmeter dual rotating retarder ellipsometry
Closest papers by meaning
- Event Fields: Capturing Light Fields at High Speed, Resolution, and Dynamic Range (CVPR 2025 · similarity 0.84)
- Active Hyperspectral Imaging Using an Event Camera (CVPR 2025 · similarity 0.82)
- Generalized Event Cameras (CVPR 2024 · similarity 0.82)
- Towards Robust Event-guided Low-Light Image Enhancement: A Large-Scale Real-World Event-Image Dataset and Novel Approach (CVPR 2024 · similarity 0.81)
- PolarScale: A Physics-Grounded Benchmark for Radiometrically Consistent RGB-to-Stokes Estimation (NeurIPS 2026 · similarity 0.80)
- Polarization Wavefront Lidar: Learning Large Scene Reconstruction from Polarized Wavefronts (CVPR 2024 · similarity 0.80)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.