Time domain deconvolution when the kernal has no spectral inverse

Deconvolution of data in the Fourier domain is sometimes hampered by the presence of zeros in the kernel. A time domain deconvolution method is proposed that can circumvent such difficulty. Added advantages are: 1) a higher signal-to-noise ratio often attainable in the deconvolved result compared to the Fourier domain solution (even with nonsingular kernels), and 2) a saving in computational speed when the unknown signal is short compared to the kernel. Numerical illustrations and application to motion blur correction are also given.