Light field calibration and 3D shape measurement based on epipolar-space

Area of Science:

• Computer Vision
• Optical Engineering
• 3D Reconstruction

Background:

• Light field cameras capture 4D light fields, complicating 3D scene point mapping.
• Existing geometrical models relate 3D points to the 4D light field, posing calibration challenges.

Purpose of the Study:

• To propose a novel epipolar-space (EPS) based geometrical model for light field cameras.
• To develop a simplified approach for 3D shape measurement using light field data.
• To provide a closed-form solution for accurate 3D reconstruction.

Main Methods:

• Developed an epipolar-space (EPS) geometrical model relating 3D scene points to a 3-parameter vector.
• Implemented a calibration method with initial linear solution and Levenberg-Marquardt nonlinear optimization.
• Validated the model using a Lytro Illum light field camera.

Main Results:

• Achieved a closed-form solution for 3D shape measurement based on the EPS model.
• Demonstrated effective calibration and 3D shape measurement performance.
• Verified results with real-world data and online datasets.

Conclusions:

• The proposed EPS geometrical model offers a simplified and effective approach to light field camera calibration.
• The method enables accurate 3D shape measurement, outperforming previous geometrical models.
• This work advances 3D reconstruction techniques using light field imaging.