Contour-Based Chain-Code Serialization for Lossless Compression of Voxelized 3D Objects

Area of Science:

• Computer Graphics
• Geometric Modeling
• Data Compression

Background:

• Voxel representations are crucial for 3D object modeling but face scalability challenges with dense grids and sparse encodings at higher resolutions.
• Existing methods like octrees and point-cloud codecs have limitations in deterministic serialization for exact binary voxel occupancy.

Purpose of the Study:

• To develop an efficient contour-based chain-code serialization method for binary voxel occupancy.
• To bridge the gap between existing voxel representations and deterministic contour-first serialization.
• To improve compression efficiency for 3D voxel data at various resolutions.

Main Methods:

• Proposed a contour-based chain-code serialization by decomposing voxel grids into 2D slices.
• Extracted foreground components and holes, encoding their contours using F4, 3OT, and F8 variants.
• Separated contour symbols from positional metadata before applying general-purpose lossless compression (zstd).

Main Results:

• The proposed method is not competitive at N=8 resolution, with octree masks achieving better compression.
• From N=16 resolution onwards, the proposed method significantly outperforms baselines, with compression gains reaching 84.37% at N=512.
• The best configuration varied: zstd + 3OT for N=8/16, and zstd + F8 for N=32 to N=512.

Conclusions:

• Contour-aware serialization combined with backend compression offers significant advantages over backend compression alone.
• The proposed method provides a scalable and efficient solution for representing and compressing 3D voxel objects at high resolutions.
• This approach advances the field of 3D data representation and compression.