Characterization of Smoothness of Multivariate Refinable Functions and Convergence of Cascade Algorithms of Nonhomogeneous Refinement Equations
Song Li
DOI: https://doi.org/10.1023/A:1027373528712
2004-01-01
Advances in Computational Mathematics
Abstract:This paper concerns multivariate homogeneous refinement equations of the form φ (x) = ∑_α∈ℤ^S a(α )φ (Mx - α ) + x∈ℝ ^s , and multivariate nonhomogeneous refinement equations of the form φ (x) = ∑_α∈ℤ^S a(α )φ (Mx - α ) + g(x)∈ℝ ^s , where ϕ=(ϕ 1 ,...,ϕ r ) T is the unknown, M is an s × s dilation matrix with m =|det M |, g =( g 1 ,..., g r ) T is a given compactly supported vector-valued function on R s , and a is a finitely supported refinement mask such that each a (α) is an r × r (complex) matrix. In this paper, we characterize the optimal smoothness of a multiple refinable function associated with homogeneous refinement equations in terms of the spectral radius of the corresponding transition operator restricted to a suitable finite-dimensional invariant subspace when M is an isotropic dilation matrix. Nonhomogeneous refinement equations naturally occur in multi-wavelets constructions. Let ϕ 0 be an initial vector of functions in the Sobolev space ( W 2 k ( R s )) r ( k ∈ N ). The corresponding cascade algorithm is given by φ _n (x) = ∑_α∈ℤ^S a(α )φ _n (Mx - α ) + g(x),∈ℝ ^s ,n = 1,2....