摘要
为进一步提高三维网格压缩算法性能,在高斯混合概率模型(GHPM)基础上,提出基于贝叶斯熵编码的局部坐标分级跳跃渐进式3D网格压缩算法。采用GHPM模型实现3D网格压缩过程的顶点创建、边沿触发器设计、面方向预测以及分级跳跃分割,实现对给定顶点的后验概率几何拓扑符号估计。基于后验概率的算术编码器进行拓扑符号编码,采用不同情景进行设计,提出渐进式的标签预测过程,实现已编码组信息的充分利用,并采用局部坐标系有效压缩几何残差。通过与对比编码器的实验验证,所提算法相对于AD、wavemesh、AAD以及RDO编码器具有更高的压缩比和压缩精度,计算性能更好。
In order to improve the computational efficiency of 3D mesh compression algorithm, with the Gauss hybrid probabilistie model (GHPM), this paper proposed the Bayesian entropy encoding based local coordinate hierarchical jump progressive for 3D mesh compression. Firstly, it used the GHPM to realize the vertex creation, the edge flip flops design, the prediction of face direction, and the ierarchicAl jumping segmentation, which realized the posterior probability estimation for geometric topological symbol of the given vertex. Secondly, based on the arithmetic encoder of the posterior probability to encode the topological symbol, this paper used the different scenarios for the design, and proposed a progressive label prediction process, to achieve the full use of encoding group information, and used local coordinate system to realize the effective compression for geometric residuals. Finally, by comparing with the experimental verification, the proposed algorithm has higher compression ratio and compression accuracy compared with four kinds of encoder which are AD, wavemesh, AAD, and RDO. And it has a better performance.
出处
《计算机应用研究》
CSCD
北大核心
2017年第10期3165-3170,共6页
Application Research of Computers
基金
国家自然科学基金资助项目(U1404602)
河南省高等学校重点科研项目(15B520006
15A520063)
河南省教育厅科学技术研究重点项目(14A520046)
关键词
贝叶斯熵编码
局部坐标
分级跳跃
网格压缩
渐进式
高斯概率模型
边沿触发
Bayesian entropy encoding
local coordinates
hierarchical jump
mesh compression
progressive
Gauss probability model
edge flip flops