Convergence rates of cascade algorithms with infinitely supported masks (Q2888797)

Let \(\varphi \in L^p({\mathbb R}^s)\) \((1\leq p \leq \infty)\) be a refinable function which satisfies the refinement equation NEWLINE\[NEWLINE \varphi (x) = \sum_{\alpha \in {\mathbb Z}^s} a(\alpha)\, \varphi (M\,x - \alpha) \quad (x\in {\mathbb R}^s), NEWLINE\]NEWLINE where the mask \(a\) is infinitely supported and the dilation matrix \(M \in {\mathbb Z}^{s\times s}\) fulfils \(\lim_{n\to \infty} M^{-n} = 0\). The cascade algorithm is the iteration scheme NEWLINE\[NEWLINE \varphi_{n+1} (x) = \sum_{\alpha \in {\mathbb Z}^s} a(\alpha)\, \varphi_n (M\,x - \alpha) \quad (n=0,1,\dots) NEWLINE\]NEWLINE with given \(\varphi_0 \in L^p({\mathbb R}^s)\).NEWLINENEWLINEIn this paper, the authors characterize the convergence rates of cascade algorithms. They extend corresponding results of \textit{R. Q. Jia} [Proc. Am. Math. Soc. 131, No. 6, 1739--1749 (2003; Zbl 1020.65109)] with finitely supported masks to the case of infinitely supported masks. It is shown that under some appropriate conditions, the cascade algorithm converges in \(L^p({\mathbb R}^s)\) with an exponential rate.