Triangle Mesh Generation from Depth Map via K-D Tree Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for converting three-dimensional image data represented by a two-dimensional image and a depth map into a polygon mesh and texture format are inefficient, leading to suboptimal results with imperfections such as holes and high computational complexity, which hinders accurate and contiguous triangle mesh generation for 3D graphics processing.

Innovation Solution

The approach involves generating a k-D tree from the depth map, positioning internal vertices within undivided regions, and forming triangles between these vertices to create a contiguous triangle mesh, allowing for efficient and accurate 3D image representation suitable for graphics processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If conventional methods are used to convert depth map to polygon mesh, then conversion can be performed, but the resulting mesh contains holes and imperfections with high computational complexity

Engineering Contradiction:
Improvemesh accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The depth map is divided into multiple triangular regions that form a structured mesh. By segmenting the conversion process into discrete triangular units with shared vertices, the method achieves both high precision (no holes) and computational efficiency (reusable vertex processing).

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Vertices are pre-positioned at specific locations (e.g., corners and centers) of triangular regions before mesh generation. This preliminary vertex placement ensures that the mesh is contiguous without holes and enables efficient batch processing of shared vertices across multiple triangles.

Inventive Principle:
Principle #10Preliminary action

2Productivity

If triangle mesh is generated for 3D graphics processing, then processing efficiency can be improved, but the conversion from depth map to mesh format is computationally intensive

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidcomputational complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

Adjacent triangular regions share common vertices, allowing the processing of a single vertex to be applied to multiple triangles simultaneously. This merging approach reduces the total number of processing operations needed, improving productivity while maintaining mesh accuracy.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The method transforms the depth map data into a structured mesh representation by changing parameters such as vertex coordinates, triangle connectivity, and depth values. This parameter transformation enables efficient GPU processing while reducing the computational burden of the conversion process through systematic vertex sharing.

Inventive Principle:
Principle #35Parameter changes

3Reliability

If conventional mesh conversion is used, then format conversion can be achieved, but the result is not contiguous and suitable for high-quality 3D display

Engineering Contradiction:
Improvemesh contiguityVSAvoidconversion complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The depth map is segmented into triangular regions with explicitly defined shared vertices and edges. This segmentation ensures that adjacent triangles are properly connected, creating a contiguous mesh structure without holes or gaps, while the systematic approach keeps conversion complexity manageable.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3353748B1Generation of triangle mesh for a three dimensional image
Publication Date: 2022.08.03 KONINKLIJKE PHILIPS NV
  • EP3353748B1 patent drawingFigure 1~2
  • EP3353748B1 patent drawingFigure 3~4
  • EP3353748B1 patent drawingFigure 5~6

AI summary

An apparatus is arranged to generate a triangle mesh for a three dimensional image. The apparatus includes a depth map source (101) which provides a depth map and a tree generator (105) generates a k-D tree from the depth map. The k-D tree representing a hierarchical arrangement of regions of the depth map satisfying a requirement that a depth variation measure for undivided regions is below a threshold. A triangle mesh generator (107) positions an internal vertex within each region of the k-D tree. The triangle mesh is then generated by forming sides of triangles of the triangle mesh as lines between internal vertices of neighboring regions. The approach may generate an improved triangle mesh that is suitable for many 3D video processing algorithms.