Research map / 3D vision / Depth and stereo
SPIDeRS: Structured Polarization for Invisible Depth and Reflectance Sensing
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.
Proposes SPIDeRS, an invisible structured-light depth and reflectance sensing method that projects per-pixel angle-of-linear-polarization patterns via a liquid crystal SLM and decodes them with a polarimetric camera.
Problem
Capturing 3D shape and reflectance of real objects without visibly altering their appearance, since conventional structured light patterns are visible and IR structured light cannot capture visible-spectrum texture or reflectance.
Approach
Modulate the angle of linear polarization (AoLP) of projected light per pixel with a TN liquid crystal SLM used without an analyzer, and extract the projected AoLP pattern from polarimetric object appearance by subtracting diffuse/ambient contributions (using an unpolarized-equivalent image), enabling depth, normals, and BRDF recovery invisibly.
Key results
- Depth reconstructions of real objects are comparable to classic intensity structured light, with mean/median errors around 0.36 to 1.18 mm across examples.
- Normal estimation on a real object yields mean/median angular error of 10.9°/9.3° for the proposed method, matching the baseline's accuracy at the same level as structured light using intensity (10.9°/9.6°).
- Recovered BRDF and normals enable plausible relighting under novel directional and colored illumination.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use SPIDeRS instead of intensity-based structured light | real-object depth reconstruction · Real objects, with and without ambient light | depth reconstruction accuracy | No clear effect |
| use SPIDeRS instead of polarization-coded structured light | real-object depth reconstruction · Real objects, including scenes with ambient light | robustness to diffuse and ambient light | Helped |
| recover normals directly from polarimetric reflection rather than deriving them from depth | SPIDeRS · Real object normal reconstruction | surface-normal detail: 10.9°/9.6° -> 10.9°/9.3° | Helped |
| use SPIDeRS to capture surface appearance during depth sensing | SPIDeRS · Real objects | simultaneous surface-texture capture | Helped |
Surprising findings
- Polarization-coded structured light [11] suffers decoding errors from low pattern contrast under diffuse and ambient light, while the proposed method remains robust.
Limitations
- Requires no ambient light for BRDF and normal estimation; ambient light is only handled for depth.
- RGB polarization projector not implemented; wavelength-dependent SLM modulation forces a single green LED for depth.
- Assumes near-coaxial specular geometry (under 20 degrees) and a single-material surface for BRDF estimation.
Experimental setup
Metrics: mean and median depth error (mm) mean and median angular normal error (degrees)
Baselines: classic intensity structured light polarization-coded structured light (Huang et al.) PCA-based normals from depth
Builds on: FMBRDF polarimetric BRDF (Baek et al.) polarization-coded structured light phase shifting Gray code
Closest papers by meaning
- PolarFree: Polarization-based Reflection-Free Imaging (CVPR 2025 · similarity 0.85)
- Fooling Polarization-Based Vision using Locally Controllable Polarizing Projection (CVPR 2024 · similarity 0.85)
- Polarization Wavefront Lidar: Learning Large Scene Reconstruction from Polarized Wavefronts (CVPR 2024 · similarity 0.84)
- SfPUEL: Shape from Polarization under Unknown Environment Light (NeurIPS 2024 · similarity 0.84)
- Glossy Object Reconstruction with Cost-effective Polarized Acquisition (CVPR 2025 · similarity 0.84)
- High-Fidelity Polarimetric Implicit 3D Reconstruction with View-Dependent Physical Representation (AAAI 2025 · similarity 0.83)
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.