A characterization of nonuniform multiwavelets using dimension function (Q1682564)

The authors study nonuniform orthonormal multiwavelets, i.e., sets \(\{\psi^\ell\}_{\ell=1}^L\subset L^2(\mathbb R)\) such that the set \[ \left \{\theta^{j/2} \psi^\ell(\theta^j \cdot - \lambda) \,:\, 1\leq \ell\leq L,\, j\in \mathbb Z,\, \lambda\in \Lambda_{r,N}\right \} \] is an orthonormal basis for \(L^2(\mathbb R)\) where \(\theta=2N\) (or more generally, \(\theta= 2Na\) with \(a\in \mathbb N\) when \(N>1\) and \(a\in \mathbb N+\{0,\frac 12\}\) when \(N=1\)) and \(\Lambda_{r,N}=\{0, \frac rN\}+ 2\mathbb Z\) where \(r\) and \(N\) are relatively prime and \(1\leq r\leq 2N-1\). In particular, they study the question when such wavelets are obtained from a nonuniform multiresolution analysis with dilation \(\theta\) and translation set \(\Lambda_{r,N}\), where the only difference compared to the classical multiresolution analyses is the requirement that there are scaling functions \(\varphi^d\), \(d=1,\ldots,D\), so that \(\{\varphi^d(\cdot -\lambda)\,:\, 1\leq d\leq D,\, \lambda \in \Lambda_{r,N}\}\) is an orthonormal basis for the subspace \(V_0\subset V_1\). It turns out that one has to have \(L=(\theta-1)D\) and one main result of the paper is that a necessary and sufficient condition for the multiwavelets \(\{\psi^\ell\}_{\ell=1}^L\) to be associated with a multiresolution analysis is that for a.e.\ \(\xi\) \[ \sum_{p\in \mathbb Z}\sum_{\ell=1}^L \sum_{j=1}^\infty \left | \widehat {\psi^\ell}\left (\theta^j\left (\xi+\frac p2\right )\right )\right |^2 = 2D, \] and \[ \sum_{p\in \mathbb Z}\sum_{\ell=1}^L \sum_{j=1}^\infty e^{i \pi \frac rN p}\left | \widehat {\psi^\ell}\left (\theta^j\left (\xi+\frac p2\right )\right )\right |^2 = 0. \]