Parallel 3D Reconstruction from Multi-Camera Depth Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Combining images from multiple cameras to generate a three-dimensional graphical reconstruction of an object is computationally intensive, leading to increased time and limiting the potential use of three-dimensional vision systems in applications that require timely image processing.

Innovation Solution

A system and method where multiple cameras capture images of an object positioned at a target location, with a console processing these images in parallel using a coarse-to-fine patch-match process on a GPU to determine depth information, optimizing it with stereoscopic and shading information, and minimizing total energy to generate a three-dimensional reconstruction efficiently.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If images from multiple cameras are combined to generate three-dimensional graphical reconstruction, then the accuracy and detail of the reconstruction is improved, but the computational time and resource requirements increase

Engineering Contradiction:
Improvereconstruction accuracyVSAvoidcomputation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent divides the computational task into parallel processing streams, where multiple processors simultaneously analyze images from different cameras. Each processor handles a specific camera's image data independently, computing depth information and graphical representations in parallel rather than sequentially, thus reducing total computation time while maintaining reconstruction accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from two-dimensional image data to three-dimensional graphical representations by computing depth information from multiple camera views. This dimensional transformation enables accurate 3D reconstruction by utilizing the spatial relationships captured across multiple camera angles, resolving the contradiction between detailed reconstruction and computational efficiency.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Manufacturing precision

If high resolution images are processed to generate detailed three-dimensional reconstructions, then the quality of the reconstruction is improved, but the computational complexity and processing time increase

Engineering Contradiction:
Improvereconstruction qualityVSAvoidcomputational complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent segments the high-resolution image processing task across multiple processors that operate in parallel. Each processor handles a portion of the computational workload independently, reducing the complexity burden on any single processing unit while collectively achieving high-quality 3D reconstruction from high-resolution input images.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary processing of high-resolution images by extracting key depth information and features before generating the final 3D reconstruction. This preliminary analysis reduces the complexity of subsequent processing steps while preserving the quality benefits of high-resolution input data.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3850834B1Generating a representation of an object from depth information determined in parallel from images captured by multiple cameras
Publication Date: 2024.11.13 META PLATFORMS TECHNOLOGIES LLC
  • EP3850834B1 patent drawingFigure 1
  • EP3850834B1 patent drawingFigure 2
  • EP3850834B1 patent drawingFigure 3

AI summary

Multiple cameras with different orientations capture images of an object positioned at a target position relative to the cameras. Images from each camera are processed in parallel to determine depth information from correspondences between different regions within an image captured by each image capture device in parallel. Depth information for images from each camera is modified in parallel based on shading information for the images and stereoscopic information from the images. In various embodiments, the depth information is refined by minimizing a total energy from intensities of portions of the images having a common depth and intensities of portions of the image from shading information from images captured by multiple cameras. The modified depth information from multiple images is combined to generate a reconstruction of the object positioned at the target position.