OctMesh improves lossless 3D mesh compression using octree structures
OctMesh: A Unified Octree-Hierarchical Framework for Lossless Triangle Mesh Compression
Computer Vision and Pattern RecognitionGraphics
Summary
Compressing 3D shapes made of triangles without losing details is hard because both the points and how they connect matter. The authors created OctMesh, which uses a special tree-like structure to organize points and connections efficiently. This method groups connections in a way that simplifies predicting their layout, letting a computer compress the mesh data more tightly. Tests show OctMesh reduces data size compared to previous methods while perfectly reconstructing the original shapes.
What this means in practice
- •For 3d graphics developers: Implement more efficient mesh storage and transmission with a method that compresses triangle meshes smaller without losing detail.
- •For video game studios: Use hierarchical mesh compression for streaming detailed 3D models progressively to improve game load times and visual quality.$Commercial implications: Enables selling games or assets with faster downloads and better visuals through improved mesh compression.
Authors
Shiyu Feng, Xihua Sheng, Lingyu Zhu, Chunyang Fu, Shiqi Wang
Abstract
Lossless triangle mesh compression must preserve both vertex positions and connectivity. Octrees support learned point cloud geometry coding and progressive refinement, but extending them to meshes requires a compatible connectivity representation. Unlike the eight occupancy decisions of a voxel, a parent edge can develop into varied child connections, making edge refinement difficult to model with a compact prediction prior. We propose OctMesh, a learned framework that codes geometry and connectivity on a shared octree hierarchy. Its key observation is that octree pooling produces parents with either one child or two to eight children. Child edges are then grouped by their endpoint parents' types and whether the endpoints share a parent. Each candidate group contains children from just one parent or two connected parents. The resulting four categories define small, fixed-shape prediction tasks: connections uniquely determined by the parent graph are inherited without bits, while three neural predictors estimate probabilities for the remaining candidates. These probabilities guide arithmetic coding of the actual edge symbols. Binarized predictions of within-parent connections provide context for predicting connections between different parents. A graph-aware parent feature extractor combines local geometry, parent connectivity and global shape. The connectivity models use dedicated weights at coarse levels and share weights at fine levels. Residual edges and a finest-level face-selection payload complete the reconstruction. On 256 frames from eight MPEG V-DMC test sequences, OctMesh losslessly recovers the finest-level vertex-coordinate, edge and unoriented face sets at an average of 7.033 bits per face, 12.8% below V-Mesh. The same hierarchical representation supports nine levels of progressive vertex-and-edge refinement.