Volumetric calibration of a plenoptic camera

Area of Science:

• Computer Vision
• Optical Engineering
• Metrology

Background:

• Plenoptic cameras capture light-field data, enabling 3D scene reconstruction.
• Current methods often rely on thin-lens assumptions, leading to inaccuracies from real-world lens distortions.
• Accurate volumetric calibration is crucial for reliable 3D scene representation.

Purpose of the Study:

• To develop and compare novel volumetric calibration techniques for plenoptic cameras.
• To address and correct for lens distortions and limitations of thin-lens models.
• To improve the accuracy and computational efficiency of 3D scene reconstruction from light-field data.

Main Methods:

• Two polynomial mapping-based volumetric calibration methods were developed: volumetric dewarping and direct light-field calibration.
• Volumetric dewarping corrects a thin-lens-based volumetric representation using post-processing.
• Direct light-field calibration integrates polynomial mapping during initial volumetric representation for object-to-sensor mapping.

Main Results:

• Both polynomial mapping methods significantly increased 3D reconstruction accuracy compared to thin-lens calibration.
• Achievable accuracy was comparable between the two polynomial mapping techniques.
• Direct light-field calibration demonstrated computational benefits by reducing processing steps.

Conclusions:

• 3D polynomial mapping functions are effective for accurate plenoptic camera volumetric calibration.
• Direct light-field calibration offers both accuracy and computational efficiency.
• The proposed methods exhibit flexibility, including for non-planar calibration scenarios.