Frame-based average sampling in multiply generated shift-invariant subspaces of mixed Lebesgue spaces (Q2233626): Difference between revisions

The article is concerned with non-uniform average sampling and reconstruction of functions in shift-invariant spaces (SIS) with multiple generators that belong to mixed Lebesgue spaces. These SIS spaces are defined as \[ V_{p,q}(\Phi_r)= \left\{\sum_{k_1\in \mathbb{Z}} \sum_{k_2\in \mathbb{Z}^d} c^T(k_1, k_2)\Phi_r(-, k_1, -k_2): c(k_1, k_2)\in \left(\ell^{p,q}(\mathbb{Z}\times \mathbb{Z}^d)^r \right) \right\},\] where \(\Phi_r=\left( \phi_1, \dots, \phi_r \right)^T \) and \(\phi_i\) belongs to the mixed Lebesgue space \( L^{p,q}( \mathbb{R} \times \mathbb{R}^d),\) \(c=(c_1, c_2, \dots, c_r)^T\) and \(c_i\in \ell^{p,q}(\mathbb{Z}\times \mathbb{Z}^d), \; i=1, 2, \dots , r,\) and \(T\) stands for the transpose. It is assumed that the family \[\left\{\phi_i (x-k_1, y-k_2) \right\}, \; i=1, 2, \dots, r, (k_1,k_2)\in \ell^{p,q}(\mathbb{Z}\times\mathbb{Z}^d)\] is a \((p,q)\) frame of \(V_{p,q}.\) The mixed Lebesgue space \( L^{p,q}( \mathbb{R} \times \mathbb{R}^d)\) is the Banach space of all functions \(f\) such that \[ \left\| f \right\|_{L^{p,q}(\mathbb{R}\times \mathbb{R}^d)}=\left\| \left\|f(x_1, x_2) \right\|_{L^q_{x_2} (\mathbb{R}^d)} \right\|_{L^p_{x_1} (\mathbb{R})} < \infty . \] The space \(\ell^{p,q}(\mathbb{Z}\times \mathbb{Z}^d)\) is defined similarly. The authors study the average sampling and reconstruction problems in the multiply generated shift-invariant subspaces \( V_{p,q}(\Phi_r)\) under three basic assumptions: \par A) The family \[\phi_i (x-k_1, y-k_2),\quad (k_1, k_2)\in \mathbb{Z}\times \mathbb{Z}^d \] is a \((p,q)\) frame for \(V_{p,q}(\Phi_r)\). \par B) The functions \(\phi_i, i=1,2, \dots, r\) are continuous and belong to the mixed Wiener amalgam space \(W(L^{1,1})\). \par C) The functions \(\phi_i\) satisfy a partial oscillation condition. The first main result shows that any function \(f\in V_{p,q}(\Phi_r)\) can be recovered from its average samples by iterative approximation projection algorithm resulting in a sequence of functions \(\left\{ f_n\right\}\) with \( f_n \rightarrow f\) uniformly. The average sampling scheme is a convolution sampling of \(f\) and another function \(\psi\in L^1(\mathbb{R}^{d+1})\) with \(\int_{\mathbb{R}}\int_{\mathbb{R}^d}\psi (x,y)dxdy =1.\) The second main result is similar, except that the average samples are given by \[\langle f, \psi_{j,k}\rangle=\int_{\mathbb{R}} \int_{\mathbb{R}^d} f(x,y)\psi_{j,k}(x,y) dx dy, \] where \(\psi_{j,k}\) are functions satisfying certain conditions. The paper is concluded with a study of the average rate of convergence in both results.