融合自阴影重建的网格模型优化

Mesh Refinement Combined with Shape From Shading

针对表面网格重建算法难以得到较好的纹理细节这一问题,提出一种融合自阴影重建(SFS)的表面网格优化算法。以表面网格模型作为初值。然后将该网格顶点投影至可视影像上获得其对应的SFS深度值。进而使用光度一致性约束算法融合SFS约束算法作为数据项、模型自身的曲率约束算法作为平滑项,共同组成网格顶点优化能量函数。最后通过梯度下降法最小化能量对网格顶点进行调整,达到优化网格模型的目的。所提算法在DTU数据集中,相比输入的网格模型精度提高了14.5%,相比现有主流表面网格优化模型算法精度提高了1.12%。实验结果表明,所提算法可以有效修正表面网格模型的纹理细节、提高网格模型精度,从而改善重建模型效果。

To solve the difficulty in obtaining good texture details using surface mesh reconstruction algorithms,a surface mesh optimization algorithm based on shape from shading(SFS) is proposed.The surface mesh model is chosen as the initial value.Then,the mesh vertex is projected onto a visual image to obtain its corresponding SFS depth value.The photometric consistency constraint algorithm is used to fuse the SFS constraint algorithm as the data item,and the curvature constraint algorithm of the model is used as the smoothing item to form the mesh vertex optimization energy function.Finally,the gradient descent method is used to minimize energy as well as adjust the mesh vertices and optimize its model.In the DTU dataset,the proposed algorithm outperforms the input mesh model by 14.5% and the existing mainstream surface mesh optimization model by 1.12%.The experimental results show that the proposed algorithm can substantially improve the texture details of surface mesh models,the accuracy of mesh models,and the effect of reconstruction models.