GPU并行加速的边折叠简化算法

Parallel processing for accelerated edge collapse simplification algorithm with GPU

针对目前串行的网格简化算法速度过慢,忽略模型细节特征的问题,提出一种基于GPU的边折叠网格简化算法。将三角网格模型划分为一系列强独立区域,采用二次误差测度计算边折叠代价,将独立区域内的所有强独立边中的最小折叠代价作为强独立区域排序值,取各个强独立区域中折叠代价最小的边并行地进行折叠操作。实验结果表明,该算法具有较快的简化速率,较好的细节特征保持能力,简化模型具有较好的拓扑连接关系和较小的几何误差。

Focusing on the issues of low speed of mesh simplification algorithms,ignoring some important detail features of original model and large geometric errors of the simplified model,an edge collapse mesh simplification algorithm based on GPU was proposed.The triangle mesh was divided into a series of super independent areas,quadric error metric was used to calculate the cost of edge collapse,the minimum cost of super independent edge collapse within the independent area was taken as the sort value of super independent area,and the super independent edge with the minimum cost was collapsed in parallel in every super independent area.Experimental results show that the proposed algorithm not only has faster simplified rate,but also maintains efficient detail features of the original model,and the simplified model has better topological connection relation and smaller geometric error.