Arbitrarily Smooth Orthogonal Nonseparable Wavelets in \R^2

accuracy compact support nonseparable orthogonal scaling functions smooth multivariate orthogonal nonseparable wavelets smoothness two-row scaling coefficients

Mathematics Subject Classification ID

Nontrigonometric harmonic analysis involving wavelets and other special systems (42C40)

Review text

So far, no general theory for the construction of smooth multivariate orthogonal nonseparable wavelets is available and only a few constructions are published. Typically, separable bases are applied in the multivariate case, although they have well-known drawbacks: very little multivariate design freedom and artificial product structure. The authors overcome this deficit by characterizing completely the set of two-row scaling coefficients that produce nonseparable orthogonal scaling functions with arbitrary high accuracy and smoothness. To be a little bit more precise, the authors treat the following problem: Let $M$ be a $2\times 2$-matrix with integer entries such that $|\text{det} M|=2$. Then, find coefficients $(c_n)_{n\in \mathbb Z^2}$ with $\sum_{n\in\mathbb Z^2} c_n =1$ such that scaling functions $\varphi:\mathbb R^2\to \mathbb R$ with \[ \varphi(x) = 2 \sum_{n\in\mathbb Z^2} c_n \varphi(Mx-n) \] possess the following properties: a) $(\varphi(\cdot - n))_{n\in\mathbb Z^2}$ are orthogonal in $L^2(\mathbb R^2)$. b) The linear span of $(\varphi(\cdot - n))_{n\in\mathbb Z^2}$ contains all polynomials of degree $r-1$ or less. The authors consider in particular the case where $M=\left (\begin{smallmatrix} 0 & 2 \\ 1 & 0\end{smallmatrix} \right)$ and the nonzero scaling coefficients are placed to two adjacent rows in the first quadrant. In a first step the scaling coefficients are explicitely derived. Afterwards it is proved that the resulting scaling functions have the desired properties. The associated wavelets $\psi$ are finally given by \[ \psi(x) = 2 \sum_{n\in \mathbb Z^2} d_n \varphi(Mx-n) \quad \text{with } d_n= (-1)^{n_1} c_{e-n}, \] where $n=(n_1,n_2)$ and $e=(1,0)$. At the end, the authors discuss generalizations to other expanding matrices: It turns out that accuracy and orthogonality depend on the sublattic type of $M(\mathbb Z^2)$, whereas the smoothness and accuracy depend on the dilation matrix. Moreover, plots of some of the scaling functions are given. In particular, the first differentiable nonseparable scaling functions together with its 48 scaling coefficients are presented. For related independent results the authors give a hint to [\textit{A. Ayache}, Rev. Mat. Iberoam. 15, No. 1, 37-58 (1999; Zbl 0923.42021)].

Recommendations

Cited in

(62)

This page was built for publication: Arbitrarily Smooth Orthogonal Nonseparable Wavelets in $\R^2$

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4255791)