Stable local bases for multivariate spline spaces (Q5945604)

The numerical application of multivariate splines requires efficient algorithms for constructing locally supported bases for the space \({\mathfrak S}_d^r (\Delta)\) of multivariate polynomial splines of smoothness \(r\geq 1\) and degree \(d\) on an arbitrary triangulation \(\Delta\) of a bounded polyhedral domain \(\Omega\subset \mathbb{R}^n\). The famous \(B\)-splines form a stable locally supported basis for \({\mathfrak S}_d^r (\Delta)\) in the case \(n=1\) for all \(d\geq r+1\). In the case \(n=2\), local bases are known for all \(d\geq 3r+2\). In the case \(n=3\), local bases are known for all \(d\geq 8r+1\). In this interesting paper, the author presents an algorithm for constructing stable local bases for \({\mathfrak S}_d^r (\Delta)\) and its superspline subspaces under the assumption \(d\geq r2^n+1\), \((n\geq 2)\). It is shown that in the multivariate case, the model smoothness conditions can be better localized than usual Bernstein-Bézier smoothness conditions.