Classification of refinable splines (Q853543)

The authors present a nice classification of refinable splines. A refinable spline is a compactly supported refinable function that is piecewise polynomial. The main result reads as follows: Let \(f\) be a compactly supported spline satisfying the refinement equation \[ f(x)=\sum_{j=0}^n c_j\,f(\lambda x - d_j) \] with \(\sum_{j=0}^n c_j= | \lambda| \), \(\lambda \in {\mathbb R}\setminus \{0\}\), and \(d_j \in {\mathbb Z}\). Then: (a) There exists an integer \(k>0\) with \(\lambda^k\in {\mathbb Z}\). (b) Let \(K>0\) be the smallest integer with \(\lambda^K\in {\mathbb Z}\). Then the compactly supported distribution solution \(\varphi\) of \[ \varphi(x) = \sum_{j=0}^n c_j\,| \lambda| ^{K-1}\, \varphi(\lambda^Kx-d_j) \] is a spline. (c) There exists a constant \(\alpha\) such that \[ f(x)=\alpha\, \varphi(x)\ast \varphi(\lambda^{-1}x)\ast\, \ldots \,\ast \varphi(\lambda^{1-K}x). \] Conversely, if the refinement equation satisfies (a) and (b), then the compactly supported distribution solution \(f\) is a spline given by the above convolutional product.