Analysis of error propagation: from raw light-field data to depth estimation.

Area of Science:

• Optics
• Computer Vision
• Image Processing

Background:

• Micro-lens arrays are fundamental to light-field imaging, capturing angular and depth information.
• Micro-lens center location accuracy directly impacts the fidelity of depth estimation.
• Existing methods lack a systematic analysis of error propagation from raw data to depth estimation.

Purpose of the Study:

• To develop and validate a method for analyzing error propagation in light-field depth estimation.
• To quantify the impact of micro-lens center positioning errors on depth accuracy.
• To demonstrate significant reduction of disparity errors through improved center positioning.

Main Methods:

• Simulated light-field images with varying parameters (aperture size, noise, distance) using backward ray tracing and Monte Carlo sampling.
• Stepwise error accumulation analysis from center positioning to sub-aperture image generation and depth estimation.
• Experimental validation using an industrial light-field camera.

Main Results:

• Disparity errors increase with greater micro-lens center positioning errors and larger defocusing distances.
• Simulations accurately model error propagation across different imaging conditions.
• Experimental results confirm significant disparity error reduction with higher center positioning accuracy.

Conclusions:

• Micro-lens center positioning accuracy is a critical factor for reliable light-field depth estimation.
• The proposed method effectively analyzes and quantifies error propagation.
• Improved micro-lens metrology can substantially enhance depth accuracy in practical light-field applications.