Principals and Pupils of Lenslet-Based Light Field Camera Calibration

Area of Science:

• Computer Vision
• Optical Engineering
• Computational Imaging

Background:

• Lenslet-based light field cameras provide rich data for 3D geometry estimation and novel view synthesis.
• Manufacturing variations in these cameras require robust calibration methods for accurate measurements.
• Existing calibration techniques have limitations in common optical configurations like telephoto and wide-field modes.

Purpose of the Study:

• To introduce a novel principal plane and pupil-centric camera model for lenslet-based light field cameras.
• To address limitations of current calibration methods in diverse optical setups.
• To provide a physically grounded and numerically stable intrinsic model for light field cameras.

Main Methods:

• Developed a principal plane and pupil-centric camera model incorporating pupil and principal plane locations and magnification.
• Derived a generally applicable intrinsic model and a 4D distortion model for complex aberrations.
• Introduced a distance-normalized image-space reprojection error for balanced spatial and angular performance.

Main Results:

• Demonstrated superior reprojection and pose estimation accuracy on commercial cameras across five focal configurations.
• Showcased stable operation across all tested optical modes.
• Achieved improved rectification performance in view synthesis and epipolar plane consistency.

Conclusions:

• The proposed model offers a robust foundation for calibrated light field processing, generalizing to various camera systems.
• It overcomes limitations of existing methods, enabling accurate calibration in telephoto and wide-field modes.
• Publicly available code and data facilitate further research and application of this robust calibration technique.