A Generic Multi-Projection-Center Model and Calibration Method for Light Field Cameras.

Area of Science:

• Computer Vision
• Optics
• Photogrammetry

Background:

• Light field cameras capture spatial and angular light ray information for single-exposure 3D reconstruction.
• Accurate 3D reconstruction heavily relies on precise intrinsic parameter characterization of light field cameras.

Purpose of the Study:

• To propose a generalized multi-projection-center (MPC) model for characterizing light field cameras.
• To develop a robust calibration algorithm for the proposed MPC model.

Main Methods:

• Introduced a 6-parameter MPC model, extending the traditional two-parallel-plane (TPP) representation.
• Derived a 3D projective transformation linking 4D ray constraints and 3D geometry to MPC coordinates.
• Developed a calibration algorithm featuring a closed-form solution and non-linear optimization for re-projection error minimization.

Main Results:

• The MPC model effectively parameterizes diverse light field imaging formations, including conventional and focused systems.
• The proposed calibration algorithm demonstrated accurate performance on both simulated and real-world scene data.
• Experimental validation confirmed the efficacy of the MPC model and calibration method.

Conclusions:

• The developed MPC model provides a comprehensive framework for light field camera characterization.
• The proposed calibration algorithm offers a reliable method for improving 3D reconstruction accuracy.
• This work advances the field of light field imaging and 3D reconstruction.