Optimal reconstruction of approximate planar surfaces using photometric stereo

Area of Science:

• Computer Vision
• Computational Imaging
• Surface Metrology

Background:

• Photometric stereo enables fast, noncontact surface reconstruction of Lambertian surfaces.
• Existing research addresses uncertainties and optimal lighting but lacks solutions for noisy real-world images.
• Optimal reconstruction from noisy data remains a challenge in surface metrology.

Purpose of the Study:

• To derive optimal surface reconstruction methods for approximate planar textured surfaces using photometric stereo.
• To address the challenge of reconstructing surfaces from noisy real images.
• To enhance the accuracy of surface reconstruction, especially for fine details.

Main Methods:

• Developed optimal surface reconstruction algorithms for photometric stereo.
• Incorporated measurable statistics of imaging errors into the reconstruction process.
• Utilized simulated and real-surface experimental studies for validation.

Main Results:

• Proposed methods demonstrate improved surface reconstruction accuracy.
• Significant enhancement observed for high-frequency height variations.
• Validation through both simulated and real-world experimental data.

Conclusions:

• The derived methods provide optimal surface reconstruction for approximate planar textured surfaces under noisy conditions.
• The approach effectively handles imaging errors by utilizing their measurable statistics.
• This work advances photometric stereo applications by enabling robust surface reconstruction from imperfect image data.