高级检索

对面向四边面片笼子中值坐标计算效率的改进

Improving the Computation Efficiency of Mean Value Coordinates for Quad Cages in 3D

  • 摘要: 针对面向四边面片笼子中值坐标(mean value coordinates for quad cages in 3D, QMVC)计算效率低的问题, 对QMVC进行改进. 改进分为两部分, 第一部分是对四边面片参数化的改进: 随着采样点与四边面片间的相对距离由远及近, 将四边面片分别视作两个三角形、有理双线性曲面和标准双线性曲面, 当四边面片被视作两个三角形时, 其顶点权重的计算速度更快, 而有理双线性曲面的引入能让两个三角形在几何上平滑地过渡为标准双线性曲面, 同时过渡的位置和速度可由预定义的参数自由控制; 第二部分是对数值积分方法的改进: 对四边面片进行剖分时采用了更高效的策略, 进一步提高计算效率. 分别对4个不同模型在同一参数以及同一模型在不同参数下做变形实验, 结果表明通过选取合适的参数, 改进方法可以平衡计算时间和变形效果. 在改进方法的默认参数下, 钉盒模型可以达到和QMVC几乎相同的变形效果, 但计算时间从7 579秒缩短至3 808秒.

     

    Abstract: To improve the computational efficiency of mean value coordinates for quad cages in 3D (QMVC), this paper presents an improved method. The improvement consists of two parts. The first part focuses on the parameterization of quad patches: as the relative distance between the sampling points and a quad patch decreases from far to near, the quad patch is successively treated as two triangles, a rational bilinear surface, and a standard bilinear surface. By treating the quad patches as two triangles, the calculation of quads’ vertex weights is significantly faster. The rational bilinear surface is introduced to allow for a smooth geometric transition from two triangles to the standard bilinear surface, with the position and rate of transition being freely controlled by predefined parameters. The second part involves an improvement in the numerical integration method: a more efficient strategy is employed for the subdivision of quad patches, further improving the computational efficiency. Experiments on four different models with the same parameters and on the same model with different parameters are conducted. The results demonstrate that by selecting appropriate parameters, the improved method can effectively balance the computational efficiency and the deformation quality. With the default parameters of the improved method, the nail box model achieves almost the same deformation effect as QMVC, but the computation time is reduced from 7 579 seconds to 3 808 seconds.

     

/

返回文章
返回