3D Mesh Encoding with Dihedral-Angle Prediction for Sharp Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D mesh compression algorithms, particularly parallelogram prediction methods, struggle with high prediction errors around sharp features in 3D engineering models due to assumptions of co-planarity and limited reference triangles, leading to inefficient geometry compression.
Innovation Solution
The method analyzes dihedral angles between triangles, clusters frequently occurring angles, and uses a representative dihedral angle for enhanced prediction, allowing for improved encoding and decoding of 3D mesh models by adjusting prediction triangles based on these clusters, reducing residuals and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If parallelogram prediction is used for geometry compression, then encoding speed is maintained, but prediction accuracy deteriorates around sharp features due to co-planarity assumptions
Solution Approach 1:
The patent segments the prediction process by dividing triangles into different categories based on their geometric properties (e.g., sharp features vs. smooth surfaces). Different prediction strategies are applied to different segments: advanced prediction with dihedral angle clustering for sharp features, and traditional parallelogram prediction for smooth surfaces, thus resolving the contradiction between speed and accuracy.
Solution Approach 2:
The patent introduces dynamic adaptation by analyzing the local geometry characteristics of each triangle and dynamically selecting the appropriate prediction mode. The system transitions between different prediction strategies based on the actual geometric properties, allowing it to maintain high speed on simple geometries while achieving high accuracy on complex sharp features.
2Device complexity
If traditional parallelogram prediction is used, then device complexity is low, but compression efficiency deteriorates due to large residuals around sharp features
Solution Approach 1:
The patent performs preliminary analysis of dihedral angles and geometric characteristics before the actual prediction process. By pre-clustering dihedral angles and identifying sharp features in advance, the system prepares optimized prediction parameters that significantly reduce residuals and improve compression efficiency without adding substantial complexity to the main prediction pipeline.
Solution Approach 2:
The patent changes the prediction parameters dynamically based on local geometry. Instead of using fixed parallelogram prediction parameters, the system adjusts prediction parameters (such as dihedral angles, reference triangle selection) according to the actual geometric properties of each region, thereby reducing prediction errors and improving compression efficiency.
3Measurement precision
If dihedral angle clustering is implemented, then prediction accuracy improves for sharp features, but encoding complexity increases
Solution Approach 1:
The patent implements self-service by having the system automatically analyze its own geometric characteristics and adaptively select prediction parameters without external intervention. The dihedral angle clustering and sharp feature detection are performed automatically based on the input mesh data, reducing the need for manual configuration and minimizing the increase in encoding complexity.
Data Source
AI summary
3D mesh models are widely used in various applications for representing 3D objects. These models are made of vertices and corresponding triangles, which can be compressed based on prediction and residuals. The present invention improves the accuracy of parallelogram prediction, particularly near sharp features. The proposed 3D mesh model encoding comprises analyzing the spatial or dihedral angles between triangles, clustering triangles with similar or equal dihedral angles, and defining a representative dihedral angle for each cluster. Triangles of each cluster are then encoded relative to individual prediction triangles having the representative dihedral angle according to the cluster. Additionally, the prediction triangle may be mirrored. An indication of the encoding mode is inserted into each vertex of the encoded bitstream. A decoder extracts the encoding mode indication, reconstructs the individual prediction triangles based on the respective representative dihedral angles and performs triangle prediction and reconstruction.