Generalized refinement equations and subdivision processes (Q1890568): Difference between revisions

There exist known relations between the convergence of the discrete subdivision processes of the form \(f^{k + 1} = Sf^ k\) where \(f^ k\) denotes a vector in \(l^ \infty (\mathbb{Z})\), \(S\) is defined as \(f^{k + 1}_ j = \sum_{i \in \mathbb{Z}}\) \(a_{j - 2i}f^ k_ i\) for \(j \in \mathbb{Z}\), and the existence of solutions of functional equations of the form \(f(x) = \sum_{j \in \mathbb{Z}} a_ j f(2x - j)\). The authors investigate analogous relations between generalizations of the above notions, namely between integral subdivision processes of the form \[ f^{k + 1} (x) = \int_{-\infty}^ \infty a(x - 2t) f^ k(t) dt = \bigl( a(2 \cdot) *f^ k \bigr) \bigl( {\textstyle {x\over 2}}\bigr) \] (here \(f^ k\) denotes a function), and integral functional equations \[ f(x) = \int_{- \infty}^ \infty a(t) f(2x - t) dt = (a*f) (2x). \] They derive a necessary and a sufficient condition for uniform convergence of subdivision schemes of both types and prove the equivalence of their weak convergence and the existence of weak compactly supported solutions to the corresponding functional equations. For some special cases stronger results are derived using probabilistic methods, as in the recent work of \textit{G. A. Derfel} [Ukr. Mat. Zh. 41, No. 10, 1322-1327 (1989; Zbl 0713.45003)]. Some interesting functions created by these schemes are also presented.