Minimum variance image interpolation from noisy and aliased samples

Interpolation of signals from a discrete set of noisy samples is addressed from a statistical standpoint. We present a reconstruction formula which is optimum in the MMSE sense, for both continuous and discrete-modeled signals, in a form which is substantially simpler than this may appear from earlier publications on the subject. We also derive the required Wienerstyle interpolation kernel to be applied in order to obtain an optimal reconstruction. For image signals, we compare the results obtained using the proposed method with the ones obtained from typical spline interpolation routines.