Research map / 3D vision / Depth and stereo
Dispersed Structured Light for Hyperspectral 3D 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.
Adds a sub-millimeter diffraction grating film in front of a conventional RGB projector to disperse structured-light patterns, enabling compact low-cost hyperspectral 3D imaging with 18.8 nm FWHM spectral accuracy and about 1 mm depth error.
Problem
Hyperspectral 3D imaging (joint depth and spectrum per pixel) is either expensive and bulky (CASSI, tunable filters) or low in spectral and depth accuracy (affordable projector-based systems).
Approach
Placing a diffraction grating film in front of a projector disperses structured-light patterns by wavelength; a dispersive projection image formation model using zero- and first-order diffraction, with binary-code depth decoding and per-pixel spectral reconstruction from scanline patterns, yields accurate hyperspectral 3D with a cheap, compact setup.
Key results
- DSL achieves 18.8 nm average FWHM spectral accuracy versus 83 nm for Li et al. on nine 10 nm bandpass filters.
- Average depth error of about 1 mm (relative) and 1.35 mm (absolute on linear stage).
- Reconstruction spans 46 channels from 430 to 660 nm at 5 nm intervals with a prototype costing around 10 USD for the grating.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use DSL instead of Li et al. [21] | Practical hyperspectral 3D imaging · Visible spectrum | spectral FWHM: 83 nm -> 18.8 nm | Helped |
| add first-order diffraction to the reconstruction (ablation) | DSL · Nine 10 nm bandpass filters | spectral FWHM: 50 nm -> 18.8 nm | Helped |
Surprising findings
- Conventional binary-code decoding remains valid despite first-order diffracted light, since zero-order light always dominates the threshold in this setup.
Limitations
- Capture takes around 10 minutes, restricting to static scenes.
- Low intensity of first-order diffraction limits working depth to under one meter.
- Only visible spectrum demonstrated.
Experimental setup
Datasets: simulated binary-code samples with random hyperspectral reflectance (106 samples) real captures of Spectralon calibration targets
Benchmarks: real-world captures with spectroradiometer ground truth ColorChecker and metameric leaves linear translation stage depth ground truth
Metrics: FWHM average depth error (mm) spectral accuracy
Baselines: Li et al. (projector-camera hyperspectral 3D) DSL without first-order diffraction conventional structured light decoding
Builds on: structured light binary coding CASSI Li et al. hyperspectral projector-camera method
Closest papers by meaning
- Dense Dispersed Structured Light for Hyperspectral 3D Imaging of Dynamic Scenes (CVPR 2025 · similarity 0.95)
- HyperGS: Hyperspectral 3D Gaussian Splatting (CVPR 2025 · similarity 0.83)
- TurboSL: Dense Accurate and Fast 3D by Neural Inverse Structured Light (CVPR 2024 · similarity 0.83)
- Active Hyperspectral Imaging Using an Event Camera (CVPR 2025 · similarity 0.83)
- Lighting Every Darkness with 3DGS: Fast Training and Real-Time Rendering for HDR View Synthesis (NeurIPS 2024 · similarity 0.82)
- 3D Convex Splatting: Radiance Field Rendering with 3D Smooth Convexes (CVPR 2025 · similarity 0.82)
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.