Predictive Position Decoding for 3D Mesh Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3D mesh compression methods are inefficient for large engineering models due to failure in effectively removing redundancy and recognizing repeating patterns across different scales, leading to suboptimal compression ratios and increased storage needs for transformation information.
Innovation Solution
The implementation of predictive probability coding during octree construction, where the probability of non-empty-child-cell configurations is predicted based on the smoothness of 3D model surfaces, allowing for adaptive entropy coding that assigns shorter codewords to high-probability symbols and longer codewords to low-probability symbols, effectively removing geometric redundancy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If components are compressed separately in multi-connected 3D models, then the compression process is simpler, but the compression ratio is reduced due to redundancy between components not being removed
Solution Approach 1:
The patent merges the compression processing of multiple connected components into a unified process. By treating the entire multi-connected model as a single processing unit rather than separate components, the system can identify and remove geometric redundancies that span across component boundaries, thereby achieving higher compression ratios while maintaining manageable process complexity through integrated handling.
Solution Approach 2:
The patent implements a universal compression framework that handles both single and multi-connected 3D models through the same processing pipeline. The system universally applies geometric feature detection, redundancy removal, and transformation encoding across different model types, eliminating the need for separate processing paths and achieving consistent high compression performance across diverse model structures.
2Productivity
If PCA of positions of vertices is used for compression, then the compression process is faster, but components with same geometry and different connectivity cannot be distinguished as repeating features
Solution Approach 1:
The patent applies local quality analysis by examining geometric features at the local component level while maintaining global context. Instead of relying solely on global PCA transformations that lose local geometric characteristics, the system performs local geometric feature detection and comparison, preserving the ability to distinguish components with identical geometry but different connectivity patterns while maintaining efficient processing through localized operations.
3Reliability
If transformation information of repeated instances is stored, then the original model can be restored accurately, but the storage requirements increase significantly
Solution Approach 1:
The patent implements a copying strategy where instead of storing complete transformation information for each repeated instance, the system stores a master definition of the repeating geometric feature and generates instances by copying and transforming this master definition. This approach dramatically reduces storage requirements while maintaining full restoration capability, as the original model can be reconstructed by applying stored transformations to the copied master features.
4Ease of manufacture
If standard entropy coding is applied to octree symbols, then the coding is simpler, but compression efficiency is reduced due to inability to exploit geometric correlation
Solution Approach 1:
The patent implements feedback-based probability prediction for entropy coding of octree symbols. The system analyzes the geometric correlation and surface smoothness from previously processed cells to dynamically predict the probability distribution of current cell symbols. This feedback mechanism allows the encoder to adapt to local geometric characteristics, significantly improving compression efficiency while maintaining coding simplicity through automated probability estimation based on geometric context.
Data Source
AI summary
A method and apparatus for position decoding of three dimensional mesh models are described including predicting a symbol probability of a non-empty-child-cell Cl,k, where Cl,k denotes the kth cell at layer l, wherein the symbol probability is estimated based on an accuracy of a fitted plane P, decoding the non-empty-child-cell responsive to the received predicted probability of the non-empty-child-cell, subdividing the non-empty-child-cell, if the non-empty-child-cell has more than one vertex, determining if there are more unprocessed non-empty-child-cells at layer l, determining if a lowest layer of non-empty-child-cells has been reached, if there are no more unprocessed non-empty-child-cells at layer l and regenerating the three dimensional mesh model, if the lowest layer of non-empty-child-cells has been reached.