The spectrum of irregular samples of a two-dimensional (2-D) signal is derived using nonharmonic Fourier series expansion. The assumption on the set of irregular samples is general except that this set should not deviate from a set of uniform samples by more than a finite amount. The spectral analysis suggests a simple method to reconstruct a 2-D signal from nonuniform samples. The accuracy of the recovery technique increases when the nonuniform samples do not deviate drastically from the uniform sampling set (less than T/π-where T is the Nyquist interval). The same analysis can be extended to n-dimensional signals.