Robust mesh deformation with salient features preservation

Triangular meshes often contain a few salient features. Traditional deformation algorithms mainly preserve the local details and volume, thus producing unnatural results. This paper proposes a robust and effective algorithm to prevent the distortion of salient features. Firstly, the salient features can be automatically extracted through saliency-based clustering and aggregation. A nonlinear energy function is then minimized to make the salient features behave rigidly to retain the shape. Finally, for the robustness of the minimization, we generate a coarse solid subspace around the input mesh, and carry out the energy minimization in this subspace. Experiments show that our algorithm can preserve the salient features and obtain visual-pleasing results.摘要创新点本文提出了一种鲁棒有效的变形算法来防止显著特征的扭曲。 首先, 基于局部显著度的聚类算法可以自动地提取出网格模型的显著特征。 然后, 通过优化一个非线性能量函数来使得这些特征进行刚性变形, 从而保持它们的形状不变。 最后, 为了保证优化过程的收敛性和稳定性, 将能量函数投影到一个稀疏的实体子空间, 并在该子空间上进行能量的优化。