2D Image to 3D Mesh Depth Mapping for Stereoscopic Rendering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for generating 3D imagery require capturing or rendering scenes from multiple perspectives, which is resource-intensive and limits the use of 2D images in producing quality 3D imagery, especially for computer-generated environments with many objects.

Innovation Solution

A computer-implemented method that maps a 2D image to a 3D mesh using depth information, allowing the image to be re-imaged from different perspectives to create stereoscopic images, which can be displayed to create the illusion of 3D depth without requiring true 3D imagery capture or rendering.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If multiple perspectives are captured or rendered to produce 3D imagery, then the quality and realism of 3D imagery is improved, but the computational resources and time required increase significantly

Engineering Contradiction:
Improvequality of 3D imageryVSAvoidcomputational efficiency
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent creates a 3D mesh copy of the scene geometry from a single 2D image, then generates multiple virtual camera perspectives by rendering this mesh copy from different viewpoints. This avoids the need to capture or render the actual scene multiple times, significantly reducing computational resources while maintaining 3D imagery quality

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent performs preliminary processing by creating a depth map and 3D mesh representation from the 2D image before generating stereoscopic views. This preliminary 3D reconstruction enables subsequent virtual camera rendering to be computationally efficient compared to rendering the full scene multiple times

Inventive Principle:
Principle #10Preliminary action

2Manufacturing precision

If a 3D camera with multiple lenses is used to capture 3D imagery, then true stereoscopic depth is achieved, but the device complexity and cost increase

Engineering Contradiction:
Improvestereoscopic depth accuracyVSAvoidcamera system complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent uses a computational intermediary process that takes a single 2D image and a depth map as input, then generates stereoscopic image pairs through 3D mesh reconstruction and virtual camera rendering. This computational intermediary replaces the need for complex multi-lens hardware while achieving similar stereoscopic depth effects

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the mechanical/optical system of multiple physical lenses and sensors with a computational system that uses image processing algorithms, depth mapping, and virtual camera rendering to generate stereoscopic imagery from a single 2D image source

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Manufacturing precision

If high-resolution rendering is performed for computer-generated environments, then image quality is improved, but the processing time and memory requirements increase

Engineering Contradiction:
Improveimage resolution qualityVSAvoidprocessing time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent creates a simplified 3D mesh copy representing the scene geometry, then renders this mesh copy from multiple virtual camera perspectives. This approach requires rendering the geometry structure once in 3D space rather than rendering high-resolution textures and lighting multiple times for each perspective, significantly reducing processing time while maintaining quality

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent separates the 3D reconstruction process into distinct components: depth map generation, 3D mesh construction, and virtual camera rendering. This segmentation allows the computationally intensive mesh construction to be performed once, with subsequent perspective generation being much faster since it only requires rendering the already-constructed mesh from different viewpoints

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3057066B1Generation of three-dimensional imagery from a two-dimensional image using a depth map
Publication Date: 2022.12.14 DREAMWORKS ANIMATION LLC
  • EP3057066B1 patent drawingFigure 1
  • EP3057066B1 patent drawingFigure 2
  • EP3057066B1 patent drawingFigure 3A~3B

AI summary

A method for generating stereoscopic images includes obtaining image data comprising a plurality of sample points. A direction, a color value, and a depth value are associated with each sample point. The directions and depth values are relative to a common origin. A mesh is generated by displacing the sample points from the origin. The sample points are displaced in the associated directions by distances representative of the corresponding depth values. The image data is mapped to the mesh such that the color values associated with the sample points are mapped to the mesh at the corresponding directions. A first image of the mesh is generated from a first perspective, and a second image of the mesh is generated from a second perspective. The first and second images of the mesh may be caused to be displayed to provide an illusion of depth.