Comparative study of deconvolution algorithms with applications in non-destructive testing

Two well-known blind deconvolution algorithms, have been developed and applied to ultrasonic signals deconvolution, in addition to three new algorithms based on the third-, fifth- and sixth-order statistics. The fullwidth at the half maximum (FWHM) was used for an assessment of the methods. The Claerbout´s (1977) parsimonious deconvolution (CLPD) needs slightly more computational power than the other methods since the update of the cross-product vector contains the logarithm of the deconvolution output. The fifth-order statistics deconvolution (FOSD) and sixth-order statistics deconvolution (SOSD) tend to give noisy results but are applicable if the signal consists of spikes with nearly the same magnitude. The CLPD results tend to have long negative spikes beside the true discovered spikes. The third-order statistics deconvolution (TOSD) is well suited for the deconvolution of ultrasonic signals