Abstract
Viterbi algorithm (VA) was originally developed to decode convolutional codes. Burkhardt and Schorb [8] used this algorithm for image restoration problem. This requires the use of a priori knowledge of the image gray level information for image modelling. An optimal estimate of the state sequence of the process observed in memoryless noise is then obtained. This leads to a nonlinear recursive filtering providing performance superior to the deconvolution method (inverse filtering). Improvement in trellis diagram leads to improvement in the computations of the probabilities.