Voxel-Based Camera Pair Selection for Free-Viewpoint 3D Reconstruction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing free-viewpoint systems face challenges in efficiently generating high-quality image or video content with a free viewpoint due to limitations in content capturing systems, such as occlusions, difficulty in deriving required information like lighting or depth, and noisy data from imprecise capturing arrangements.

Innovation Solution

A data structure is pre-computed to facilitate 3D reconstruction by generating a map of overlapping camera frustums, determining suitable camera pairs for each voxel based on epipolar diagonals and principal points, and assigning weightings to these camera pairs for efficient depth data extraction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If a greater number of cameras are included to mitigate problems associated with occlusions and depth data extraction, then the quality and coverage of 3D reconstruction is improved, but the device complexity and cost increase significantly

Engineering Contradiction:
Improve3D reconstruction qualityVSAvoidnumber of cameras
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent pre-computes a data structure that stores camera pair identifiers and weighting values before actual 3D reconstruction occurs. This preliminary organization of camera relationships allows the system to efficiently select appropriate camera pairs during reconstruction without needing to process all possible camera combinations in real-time, thereby achieving high-quality reconstruction with fewer cameras

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent divides the scene into multiple voxels and associates different camera pairs with different voxel regions based on their depth coverage characteristics. This segmentation allows each camera pair to be optimized for specific spatial regions, improving overall reconstruction quality without requiring every camera to capture the entire scene

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If more processing is applied to address occlusions and derive depth information, then the accuracy of free-viewpoint content is improved, but the processing time and computational resources increase

Engineering Contradiction:
Improvedepth data accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system pre-computes and stores a data structure containing camera pair identifiers, weighting values, and depth coverage information before actual 3D reconstruction. This preliminary processing organizes all necessary depth extraction information in advance, allowing the reconstruction process to simply query and apply pre-determined camera pair selections rather than performing complex analysis during real-time rendering

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent creates a data structure that copies and stores essential camera relationship information (pair identifiers, weightings, depth coverage) in an optimized format. This copied representation allows rapid access during reconstruction without needing to re-analyze original camera images, significantly reducing processing time while maintaining accuracy

Inventive Principle:
Principle #26Copying

3Reliability

If all camera pairs are processed to extract depth data for every voxel, then complete scene coverage is achieved, but the computational complexity and processing time increase significantly

Engineering Contradiction:
Improvescene coverage completenessVSAvoidreconstruction efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent segments the scene into voxels and assigns specific camera pairs to specific voxel regions based on pre-computed depth coverage analysis. This segmentation ensures that each voxel is processed by the most suitable camera pair, achieving complete scene coverage while avoiding the inefficiency of processing all camera pairs for all voxels

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system applies different camera pairs to different spatial regions (voxels) based on local depth coverage characteristics. Each voxel receives processing from camera pairs that are locally optimized for that region, ensuring reliable coverage while improving overall reconstruction efficiency through localized optimization

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3790276B1Free-viewpoint method and system
Publication Date: 2025.06.18 SONY INTERACTIVE ENTERTAINMENT LLC
  • EP3790276B1 patent drawingFigure 1~2
  • EP3790276B1 patent drawingFigure 3~4
  • EP3790276B1 patent drawingFigure 5

AI summary

A method of generating a 3D reconstruction of a scene, the scene comprising a plurality of cameras positioned around the scene, comprises: obtaining the extrinsics and intrinsics of a virtual camera within a scene; accessing a data structure so as to determine a camera pair that is to be used in reconstructing the scene from the viewpoint of the virtual camera; wherein the data structure defines a voxel representation of the scene, the voxel representation comprising a plurality of voxels, at least some of the voxel surfaces being associated with respective camera pair identifiers; wherein each camera pair identifier associated with a respective voxel surface corresponds to a camera pair that has been identified as being suitable for obtaining depth data for the part of the scene within that voxel and for which the averaged pose of the camera pair is oriented towards the voxel surface; identifying, based on the obtained extrinsics and intrinsics of the virtual camera, at least one voxel that is within the field of view of the virtual camera and a corresponding voxel surface that is oriented towards the virtual camera; identifying, based on the accessed data structure, at least one camera pair that is suitable for reconstructing the scene from the viewpoint of the virtual camera, and generating a reconstruction of the scene from the viewpoint of the virtual camera based on the images captured by the cameras in the identified at least one camera pair.