Entropy interpretation of maximum‐likelihood deconvolution

Entropy concepts are embodied in the maximum‐likelihood deconvolution (MLD) method, just as they are in minimum‐entropy (MED) methods. MLD asymptotically minimizes Shannon’s entropy of the reflectivity sequence (which, in MLD, is modeled as a Bernoulli‐Gaussian random sequence). Study of maximum‐ likelihood detection and estimation of the reflectivity sequence reveals that MLD is embodied in the general framework of the “adaptive” MED methods. Comparisons based on similarities and differences between MLD and various existing MED techniques show that MLD is robust due to explicit inclusion of noise in its statistical model.