Anisotropic multiscale systems on bounded domains (Q2178832)

Multiscale system of wavelets designed for $L^2(\mathbb{R}^2)$ or $L^2(\Omega)$ ($\Omega\subset\mathbb{R}^2$) could provide optimally sparse approximations of functions governed by point singularities, while anisotropic multiscale systems such as curvelets and shearlets are capable of optimally approximating certain classes of multivariate functions with singularities along hypersurfaces such as images or the solutions of various types of partial differential equations (PDE). However, the currently anisotropic multiscale systems are initially designed for $L^2(\mathbb{R}^2)$, whereas most PDEs are defined on a bounded domain $\Omega\subset\mathbb{R}^d$, so the development of efficient PDE solvers crucially depends on the construction of anisotropic systems on bounded domains satisfying various boundary conditions. This paper proposes a solution to this, where a hybrid shearlet-wavelet multiscale system on a bounded domain $\Omega\subset\mathbb{R}^2$ is constructed. The main idea is to only choose the shearlets with support fully contained in $\Omega$ from a compactly supported shearlet frame for $L^2(\mathbb{R}^2)$ to provide the optimal approximation rate for anisotropic structures, and to carefully select boundary-adapted wavelets from a boundary wavelet frame for $L^2(\Omega)$ for handling the boundary. This proposed hybrid shearlet-wavelet systems consisting of shearlets and boundary-adapted wavelets form frames for the Sobolev spaces $H^s(\Omega), s\in\mathbb{N}\cup\{0\}$, and yield optimally sparse approximations of cartoon-like functions on boundary domains $\Omega\subset\mathbb{R}^2$. This work not only has theoretical value but also is significant in applications such as imaging science and the numerical analysis of partial differential equations.