A novel approach to image enhancement using the generalized maximum likelihood criterion

The generalized maximum-likelihood algorithm is a powerful recursive scheme for waveform estimation. This algorithm uses the steepest descent algorithm, and tends toward the maximum likelihood estimate of the waveform. The convergence and sensitivity of this estimator as applied to a synthetic image are studied. This algorithm provides an excellent means of filtering and identifying objects hidden in high noise. Its effectiveness is largely dependent on two parameters, the original image´s covariance function and the convergence coefficient. Paramount to good filtering is the covariance function. A large deviation of the estimated covariance function from its actual value will result in ambiguous object identification and a near monotone estimate. Simulation experiments to support the validity of the results are presented