Snapshot 3D and texture imaging with structured illumination.

Area of Science:

• Computer Vision
• Computational Imaging
• 3D Reconstruction

Background:

• Accurate 3D surface geometry and texture reconstruction are crucial for various applications.
• Existing methods often require multiple captures or complex setups.

Purpose of the Study:

• To develop a single-shot 3D and texture imaging method.
• To enable joint recovery of 3D surface geometry and RGB reflectance.
• To leverage physics-guided, unsupervised learning for improved reconstruction.

Main Methods:

• A calibrated projector-camera system was modeled using a differentiable image formation model.
• Physics-guided, unsupervised learning was employed.
• Structured illumination was used for data capture.

Main Results:

• The method successfully recovered 3D surface geometry and RGB reflectance from a single image.
• Experimental results showed good agreement with conventional structured-light reconstruction.
• Millimeter-level depth accuracy was achieved, consistent with triangulation-based analysis.

Conclusions:

• The demonstrated snapshot method provides a promising solution for 3D and texture imaging.
• Potential applications include autonomous vehicle navigation and industrial metrology.
• The physics-guided, unsupervised learning approach enhances reconstruction accuracy.