Light Field Photometric Calibration via Sub-Aperture Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Light field cameras suffer from photometric distortion due to vignetting effects, making it difficult to obtain accurate per-pixel disparity maps and estimate vignetting parameters, which obstruct view interpolation and depth estimation in 4D light field images.
Innovation Solution
A photometric calibration method that involves obtaining a target image and a reference image, detecting feature points, constructing a triangular mesh, determining an affine transform, and calculating a photometric calibration ratio to correct intensity distortions in the target image, resulting in a distortion-free image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional photometric calibration methods are used, then the vignetting center assumption simplifies the calibration process, but the method becomes inadequate for light field images where vignetting parameters are image-dependent and coupled
Solution Approach 1:
The patent segments the light field image into multiple sub-aperture images, each with its own vignetting characteristics. By processing each sub-aperture image separately and then combining the results, the method avoids the need for a single global vignetting center assumption while maintaining manageable computational complexity.
Solution Approach 2:
The patent changes the calibration approach from assuming fixed vignetting parameters to estimating image-dependent parameters for each sub-aperture image. This parameter change allows the calibration to adapt to the specific characteristics of each view in the light field, improving accuracy without requiring overly complex global models.
2Ease of manufacture
If photometric distortion is not eliminated, then the captured light field images maintain their original appearance, but view interpolation and depth estimation are obstructed
Solution Approach 1:
The patent applies photometric calibration as a preliminary step before view interpolation and depth estimation. By correcting the photometric distortion in advance, the subsequent processing steps can proceed with accurate data, ensuring reliable depth estimation without adding complexity to the capture process.
3Device complexity
If a single moving camera is used to capture light field, then the equipment requirement is simple, but the capture speed is slow and only works in controlled environments
Solution Approach 1:
The patent uses a camera array where multiple cameras simultaneously capture different views of the scene, effectively copying the capture function across multiple devices. This parallel approach enables fast light field capture in uncontrolled environments while maintaining relatively simple individual camera systems.
4Productivity
If numerous cameras are put into a camera array, then the dynamic light field can be captured at video-rate, but the system becomes cumbersome and expensive
Solution Approach 1:
The patent segments the light field capture task across multiple cameras in an array, with each camera capturing a subset of views. By dividing the overall capture function and processing each segment independently through the proposed calibration method, the system achieves video-rate capture while managing complexity through modular processing.
Data Source
AI summary
A photometric calibration method includes: obtaining a target image and a reference image by a light field camera and transforming the reference image into a revised target image corresponding to the target image; calculating a photometric calibration ratio according to a function for correlating the target image and the revised target image by using a numerical algorithm; and the intensity of the target image is divided by the photometric calibration ratio to obtain a calibrated target image. A light field of a scene is formed by all of the calibrated target image so as to improve the quality of the light field.