Subdivision algorithms for the generation of box spline surfaces (Q1067751): Difference between revisions

This paper presents a unified approach to subdivision algorithms for generating box spline surfaces. Discrete box splines \(b_ H(\cdot | m)\), a key to the subdivision algorithm, are first studied in some detail. The relation \[ B(x-\beta | X)=\sum_{\alpha \in L(XH)}b_ H(\alpha -\beta | X)B(x-\alpha | XH), \] where \(\beta\in L(X)\), between discrete box splines and box splines is established. In addition, it is shown that \(b_ H(\alpha | H)=\det H\) if \(\alpha =XH\beta\), and the i-th component \(\beta_ i<h_ i^{-1}\) where \(H_ i=diag(...h_{i-1},h_ i,...)\), \(i=1,...,| X|\), and \(b_ H(\alpha | H)=0\) otherwise. For \(| X| >s\) and \(1\leq i\leq n\), the discrete box splines \(b_ H(\cdot | X)\) also can be written as \[ b_ H(\alpha | X)=p_ i^{-1}\sum^{p_ i-1}_{j=0}b_{H_ i}(\alpha -jh_ ix^ i| X\setminus \{x^ j\}), \] \(\alpha\in L(XH)\). Discrete box splines give good estimates to box splines via the inequality \(| B(\alpha | X)-(\det YH_ Y)^{-1}b_ H(\alpha | X)| \leq \gamma \| H\|\) where \(\alpha\in L(XH)\), \(\gamma\) is a constant depending only on X, \(Y\subset X\) and \(\| H\| =\max \{h_ i:i=1,...,| X| \}\).