SSVDAG*: Efficient Volume Data Representation Using Enhanced Symmetry-Aware Sparse Voxel Directed Acyclic Graph

Volume data, which are mostly formed as the regular grids of scalar or vector values, can represent significant amount of data, when for instance 16K
³
grid of 8b/vox dataset consumes 4TB of space not only on secondary storage, but also in operating memory of computer or graphics card. That is why compact representations of volume data, based on lossless compression, are designed. Family of data structures based on octrees was developed, including Sparse Voxel Octree (SVO), Sparse Voxel Directed Acyclic Graph (SVDAG) and Symmetry-aware Sparse Voxel Octree (SSVDAG). The paper introduces SSVDAG* data structure, which represents further enhancement of the SSVDAG, and provides better compression results without significant increase of decompression overhead.