Plenoptic Camera 3D Reconstruction Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing three-dimensional reconstruction methods using plenoptic cameras have low resolution due to the limitation of obtaining only one depth coordinate per optical element, resulting in poor three-dimensional reconstruction quality.
Innovation Solution
A method that determines three-dimensional coordinates of sampling points and uses a reconstruction grid to calculate a dissimilarity index based on pixel intensity dispersions, allowing for higher resolution three-dimensional reconstruction by associating multiple pixels with each sampling point and distributing points in a grid to enhance depth coordinate determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional epipolar image reconstruction is used with one segment per macro-pixel, then the process is simple, but the three-dimensional reconstruction resolution is low
Solution Approach 1:
The invention divides each macro-pixel into multiple virtual segments (first segment, second segment, etc.), where each segment corresponds to different angular ranges of light rays. This segmentation allows multiple depth coordinates to be obtained per macro-pixel, thereby improving three-dimensional reconstruction resolution without requiring additional physical optical elements
Solution Approach 2:
The invention introduces an angular dimension by dividing macro-pixels into segments based on different light ray angles. This transforms the single-depth-coordinate limitation into a multi-depth-coordinate system by adding angular discrimination, effectively increasing the information dimensionality available for three-dimensional reconstruction
2Measurement precision
If the number of optical elements is increased to improve resolution, then the three-dimensional reconstruction quality improves, but the device complexity and cost increase
Solution Approach 1:
The invention creates virtual copies of depth measurement capability through computational segmentation. Instead of physically multiplying optical elements, it generates multiple virtual measurement channels (segments) from each physical macro-pixel by processing different angular ranges of light rays, achieving enhanced resolution without increasing hardware complexity
3Measurement precision
If multiple pixels are associated with each sampling point, then the resolution increases, but the calculation complexity increases
Solution Approach 1:
The invention performs preliminary association between pixels and sampling points during the calibration phase, storing these relationships in advance. During actual three-dimensional reconstruction, this pre-established mapping allows efficient retrieval and processing of multiple pixels per sampling point without real-time computational overhead, thereby improving resolution while managing calculation complexity
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This method achieves a resolution at least fifty times greater than the array of optical elements, enabling high-resolution three-dimensional reconstruction of surfaces extending beyond or within the object plane, without being limited by the number of optical elements.
Implementation Method 1
Each optical element 121 receives light rays from the surface of interest 200, and having passed through the input optic 110. After passing through an optical element 121, the light rays propagate to a matrix optical sensor 130.
Implementation Method 2
The 130 matrix optical sensor is a photosensitive sensor, for example of the CCD sensor type, configured to convert an incident flow of photons into an electrical signal, to form an image.
Implementation Method 3
The said selection of pixels, associated with the same optical element, forms a macro-pixel 131. Light rays originating from the same point on the surface of interest 200 propagate through the input optical system 110, and the matrix of optical elements 120, to different pixels of the matrix optical sensor 130.
Data Source
Figure 1A~1C
Figure 2~3
Figure 4
AI summary
The invention relates to a method for three-dimensional reconstruction of a surface of interest (200), using a plenoptic camera (100), comprising the following steps: - determining the three-dimensional coordinates of a series of points of the object field of the plenoptic camera, referred to as sampling points (Pj); - determining calibration data, associating at least two pixels (pk) of the optical sensor matrix with each sampling point (Pj); - defining a reconstruction grid, each point of which is associated with one or more sampling points (Pj); - acquiring (401) at least one image of the surface of interest using the plenoptic camera (100); - from the calibration data and from the image of the surface of interest, calculating (402), for each point of the reconstruction grid, the dissimilarity index value (Cn), as a function of one or more deviations, each deviation representing a separation between the intensity values sampled, over the image of the surface of interest, by the pixels associated with one of the corresponding sampling points (Pj); - determining (403) a three-dimensional distribution of the points of the reconstruction grid, each assigned with the dissimilarity index value (Cn) of same; - and three-dimensional reconstruction (404) of the surface of interest.