An efficient trivariate algorithm for tetrahedral Shepard interpolation (Q2302406)

Suppose a function \(f\) is evaluated at nodes \(X_n=\{x_1,\dots,x_n\}\) of \(\mathbb{R}^3 \) and denote the corresponding values by \(F_n=\{f_1,\dots,f_n\}\). Let \(T=\{t_1,\dots,t_n\}\) be a set of tetrahedra with vertices in \(X_n\). Denoted by \(V_j =\{ x_{j1}, x_{j2}, x_{j 3}, x_{j 4}\}\) the set of vertices \(t_j\), for \(j = 1,\dots, m\), it is assumed that the set \( \{V_j\}_{j = 1,\dots, m} \) constitutes a cover \(X_n\), that is \[ \bigcup_{j=1}^{m} V_j=X_n. \] Each tetrahedron \(t_j\) is associated with the set of barycentric coordinates of a point \(x\in \mathbb{R}^3 \), that is \(\lambda_{j,j_1}(x),\lambda_{j,j_2}(x),\lambda_{j,j_3}(x),\lambda_{j,j_4}(x).\) The linear polynomial \(L_j (x)=\sum_{k=1}^{4}\lambda_{j,j_k} f_{j_k}\) will interpolate the data at the vertices of the tetrahedra \(t_j\). Let \[ B_{\mu,j}(x)=\frac{\prod_{l=1}^{4} \|x-x_{j_l}\|_2^{-\mu}}{\sum_{k=1}^m \prod_{l=1}^{4} \|x-x_{k_l}\|_2^{-\mu}}, \quad j=1,\dots,m, \; \mu >0, \] where \(\|\cdot\|_2\) is the Euclidean distance in \(\mathbb{R}^3\). The tetrahedral Shepard method is then defined by \[ \mathcal{T}_\mu[f](x)=\sum_{j=1}^m B_{\mu,j}(x) L_j(x). \] In this paper, properties of the tetrahedral basis function \(B_{\mu,j}(x)\) are proved. It is shown that the operator \(\mathcal{T}_\mu\) is an interpolation operator. Then a formula for the error rate is proved and the convergence is analyzed. The algorithms that are used to implement the tetrahedral Shepard method are proposed. Moreover, the performance of the trivariate interpolation algorithm is illustrated.