Polarization 3D reconstruction based on multi-exposure fusion.

Area of Science:

• Computer Vision
• Optical Metrology
• 3D Imaging

Background:

• Polarization-based 3D reconstruction provides high precision and cost-effectiveness.
• Existing methods often neglect noise suppression during image acquisition, impacting reconstruction quality.

Purpose of the Study:

• To develop an improved polarization-based 3D reconstruction method.
• To address noise and distortion issues in image acquisition for enhanced 3D accuracy.

Main Methods:

• A pixel-wise multi-exposure fusion approach was employed.
• Optimal responses within the linear camera curve were selected to minimize photon noise and nonlinear distortions.
• A highlight correction strategy was implemented to reduce specular-gradient errors.

Main Results:

• The proposed method effectively suppresses Gaussian and systematic noise.
• Polarization accuracy was enhanced, leading to more precise surface normals.
• 3D reconstruction details were significantly clearer, with a 14.3% reduction in surface normal RMSE compared to traditional methods.

Conclusions:

• The developed method offers superior noise suppression for polarization-based 3D reconstruction.
• It achieves higher accuracy in surface normal estimation and overall 3D model quality.
• The technique provides a more robust and precise solution for detailed 3D surface reconstruction.