GUSLO: General and Unified Structured Light Optimization

TL;DR

GUSLO advances structured light 3D reconstruction by coupling single-shot geometric calibration with artifact-aware photometric adaptation in a unified framework. It achieves this through a Delaunay-barycentric global matching that turns sparse projector-camera matches into dense correspondences, and a physics-guided photometric transfer using PCNet together with a Weighted-Constrained Thin-Plate Spline DeArtifact module. The approach is validated across binary, speckle, and color-coded SL patterns, showing consistent improvements in decoding accuracy and cross-encoding robustness while supporting an open benchmarking system for reproducible SL research. The methods enable reliable SL reconstruction under challenging lighting and textures, with practical implications for industrial inspection and cultural heritage digitization.

Abstract

Structured light (SL) 3D reconstruction captures the precise surface shape of objects, providing high-accuracy 3D data essential for industrial inspection and cultural heritage digitization. However, existing methods suffer from two key limitations: reliance on scene-specific calibration with manual parameter tuning, and optimization frameworks tailored to specific SL patterns, limiting their generalizability across varied scenarios. We propose General and Unified Structured Light Optimization (GUSLO), a novel framework addressing these issues through two coordinated innovations: (1) single-shot calibration via 2D triangulation-based interpolation that converts sparse matches into dense correspondence fields, and (2) artifact-aware photometric adaptation via explicit transfer functions, balancing generalization and color fidelity. We conduct diverse experiments covering binary, speckle, and color-coded settings. Results show that GUSLO consistently improves accuracy and cross-encoding robustness over conventional methods in challenging industrial and cultural scenarios.

Figures (15)

• Figure 1: Robust 3D reconstruction with our framework. (a–b) Overexposure disrupts pattern coding, leading to missing geometry. (c–d) Our projection compensates illumination loss, enabling complete reconstruction. (e–f) Surface texture degrades color-coded decoding, causing partial mesh loss. (g–h) Texture-aware chromatic optimization restores decoding and improves reconstruction completeness (highlighted in red).
• Figure 2: Model details of our GUSLO. Stage-I filters erroneous codes based on relative positions in the camera plane, then applies Delaunay triangulation and barycentric interpolation to extend discrete matches into continuous global correspondences between camera and projector pixels. In Stage-II, decoding artifacts are corrected via the DeArtifact Module, implemented using our WC-TPS method trained with ShadingNet data. The refined results are then processed by PCNet for photometric compensation, yielding optimized structured light patterns for high-precision 3D reconstruction.
• Figure 3: Comparison with the method of global encoding using Gray code (GC) and WarpingNet.
• Figure 4: Visual comparison of optimized structured light patterns under different conditions: darkroom (rows 1 & 4), multi-lighting (rows 2, 3, 5), Lambertian (rows 2 & 3), non-Lambertian (rows 1, 4, 5), textured (rows 1, 4, 5), and textureless surfaces (rows 2 & 3). Our optimized patterns yield more complete and accurate geometry across all scenarios.
• Figure 5: Global matching results with and without the error code (EC) filter. The fourth column is an enlarged view of the erroneous matches, with red dots indicating incorrect matches, green dots indicating correct matches, and blue dots indicating unmatched points.