Multivariate generalized sampling in shift-invariant spaces and its approximation properties (Q1018363)

The classical Whittaker-Shannon-Kotel'nikov sampling theorem states that any function \(f\) band-limited to \([-1/2, 1/2]\), that is, \[ f(t)= \int^{1/2}_{-1/2}\widehat f(w) e^{2\pi itw}\,dw,\quad t\in\mathbb{R}, \] may be reconstructed from its sequence of samples \(\{f(n): n\in\mathbb{Z}\}\) as \[ f(t)= \sum^\infty_{n=-\infty} f(n)\text{sinc}(t- n),\quad t\in\mathbb{R}, \] where sinc denotes the cardinal sine function, \(\text{sinc\,}t:= (\sin\pi t)/\pi t\). Thus, the Paley-Wiener space of functions band-limited to \([-1/2,1/2]\) is generated by the integer shifts of the sinc function. Under appropriate hypotheses, any multivariate function in a shift-invariant space can be recovered from its samle at \(\mathbb{Z}^d\). However, in many common situations the available data are samples of some convolution operators acting on the function itself. This leads to the problem of multivariate generalized sampling in shift-invariant spaces. In the present paper, an \(L^2(\mathbb{R}^d)\) theory involving the frame theory is exhibited, where \(d> 1\) is an integer.