On the existence of solutions to non-parametric fairing problems (Q1125310)

Surface fairing algorithms in computer-aided design are often based on the minimization of a functional which is independent of the parametrization and includes the principal curvatures \(\kappa_1\), \(\kappa_2\). However, it is not known whether or under which conditions the underlying variational problems possess a solution. For surfaces \(\Phi\), which may be written as graph of a function \(s:\Omega\to \mathbb{R}\) over a bounded set \(\Omega\) in \(\mathbb{R}^2\), the authors consider a class of such functionals, namely \[ F_{\psi,p}:= \int_\Phi (\kappa^2_1+ \kappa^2_2)^p d\mu+ \psi(a(\Phi)) \] with \(1<p<\infty\). Here, \(d\mu\) denotes the surface measure on \(\Phi\), \(a(\Phi)\) is the surface area and \(\psi\) a nonnegative lower-semicontinuous function. Given a reference function \(s_0\in H^{1,\infty}(\Omega)\) and a constant \(M> 0\), it is proved that the variational problem \(F_{\psi,p}\to \min\) has a minimizer in a class of surfaces whose describing functions \(s\) lie in the set \(\{s\in H^{2,2p}(\Omega):\|s- s_0\|_{1,\infty}\leq M\}\). As may be the case in practical applications, the set of competing surfaces can be restricted by additional constraints involving point and gradient data.