Regular multivariate sampling and approximation in \(L^p\) shift-invariant spaces (Q536263)

The classical Whittaker-Shannon-Kotel'nikov sampling formula has its counterpart in \(d\) dimensions, according to which any function \(f\) band-limited to the \(d\)-dimensional cube \([-1/2,1/2]^d\), i.e., \[ f(t)= \int_{[-1/2,1/2]^d}\widehat f(x) e^{2\pi ix^Tt}\,dx,\quad t\in\mathbb{R}^d, \] may be reconstructed from its sequence of samples \(\{f(\alpha):\alpha\in \mathbb{Z}^d\}\) as follows: \[ f(t)= \sum_{\alpha\in\mathbb{Z}^d} f(\alpha)\text{sinc}(t_1- \alpha_1)\cdots\text{sinc}(t_t- \alpha_d),\quad t= (t_1,\dots, t_d)\in\mathbb{R}^d. \] Although Shannon's sampling theory has had an enormous impact, its implementation has a number of problems. The band-limiting operation generates Gibbs oscillations, and the sine function has a very slow decay, which makes computation in the signal domain very inefficient. In the present paper, some results are proved about regular multivariate generalized sampling in the \(L^p\) setting, where \(1\leq p\leq\infty\). These sampling formulas include regular samples of the function, its derivatives and/or some filtered versions of the function itself taken at a lattice of \(\mathbb{R}^d\). Approximation schemes using these generalized sampling formulas are also included.