Multiview correlation is based on the use of explicitly parameterizing surfaces whose deformation is to be measured. In the present analyses, the surfaces are modeled as triangular facets (for local approaches), and with finite-element meshes (for global approaches) using a linear interpolation between nodal displacements. Displacement uncertainty due to the presence of noise in the images is considered. The noise floor levels of the latter are compared for both local and global cases with two different sets of pictures. A priori estimates based on the Hessian of the registration procedure are derived analytically, and a simplified expression is provided that is shown to be in very good agreement with their a posteriori evaluations. The uncertainty difference between local and global approaches is interpreted from their different spatial resolutions.

中文翻译：

---

多视图相关方案的位移不确定性

多视图相关基于使用显式参数化要测量变形的表面。在目前的分析中，表面被建模为三角形面（对于局部方法），并使用有限元网格（对于全局方法）使用节点位移之间的线性插值。考虑了由于图像中存在噪声而导致的位移不确定性。后者的本底噪声水平在局部和全局情况下与两组不同的图片进行比较。基于 Hessian 的配准程序的先验估计是通过分析推导出来的，并提供了一个简化的表达式，它表明与它们的后验评估非常一致。