Light field camera calibration and point spread function calculation based on differentiable ray tracing.

Area of Science:

• Optics and Photonics
• Computational Imaging

Background:

• Light field (LF) cameras with micro-lens arrays (MLAs) face challenges with aberrations, complicating hardware calibration and point spread function (PSF) calculation.
• Existing methods struggle with generalization, especially under severe aberration conditions.

Purpose of the Study:

• To develop an end-to-end imaging model for LF cameras that accounts for aberrations.
• To enable precise calibration of intrinsic hardware parameters and accurate PSF calculation.
• To create a generalized solution for various LF camera applications.

Main Methods:

• An end-to-end imaging model based on differentiable ray tracing was developed.
• A projection method was incorporated to reduce memory overhead associated with MLA imaging.
• A hybrid calibration approach using a genetic algorithm followed by gradient-based optimization was employed.

Main Results:

• The proposed model successfully renders LF images (PSFs) from point source locations.
• The method effectively calibrates LF camera parameters even with significant aberrations.
• The aberration-aware approach demonstrates improved generalization capabilities.

Conclusions:

• The differentiable ray tracing model provides an effective solution for aberration-aware LF camera imaging and calibration.
• This approach is promising for enhancing LF camera performance in challenging imaging environments.
• The developed model offers a generalized framework for LF camera applications requiring accurate PSF estimation.