The USC Andrew and Erna Viterbi School of Engineering USC Signal and Image Processing Institute USC Ming Hsieh Department of Electrical Engineering University of Southern California

Technical Report USC-SIPI-376

“A Geometry-Centric Approach to Progressive 3D Graphics Coding”

by Jingliang Peng

August 2006

With increasing availability and complexity of 3D models, it is _essential to compress 3D model data efficiently. In this research, we _have made a comprehensive survey of current 3D triangular mesh coding _techniques, and proposed two progressive lossless 3D_triangular mesh coders. _ First, a survey was made, which classified, described and compared prior 3D triangular mesh coders. In addition, several future directions in the field of 3D model compression were discussed. _ Then, an octree (OT)-based progressive lossless 3D triangular mesh encoder was proposed. This OT-based mesh coder encodes 3D triangular meshes with arbitrary topological structures. For a given mesh, an octree structure is first built, which organizes the input vertices into cubic cells. Then, the octree is traversed from the root gradually to the leaves, and the intermediate mesh quality is progressively refined by cell subdivisions. For each cell subdivision, both local geometry and connectivity changes are encoded, where the connectivity coding is guided by the geometry coding. Furthermore, prioritized cell subdivision is performed in the tree front to provide better rate-distortion (R-D) performance. Based on our experiments, the proposed mesh coder significantly outperforms the prior art, the kd-tree algorithm. _For instance, the improvement in geometry coding efficiency is around 10%~20%, but may go up to 50%~60% for meshes with highly regular geometry data and/or tight clustering of vertices._ Finally, an iterative-vertex-set-split (IVSS)-based progressive lossless 3D triangular mesh encoder was proposed, which also encodes 3D triangular meshes with arbitrary topological structures. Given a 3D mesh, all its vertices form an initial vertex set which is then iteratively split. At any intermediate stage of the iterative vertex set split process, the representatives of all current vertex sets and their interconnection form an approximation to the original mesh. For each vertex set split, the offsets of child representatives from the parent and the associated connectivity changes are encoded. With delicate vertex set split control, adaptive offset quantization and effective prediction, significant R-D improvement has been observed over the OT-based coder; i.e., distortion reduction of 50%~80% has been observed at the same bitrates, which makes the IVSS-based coder a practical solution in streaming applications.


This report is not currently available in PDF format for downloading. Contact the Signal and Image Processing Institute for information on its availability.