IMAGE-RESTORATION USING AN ESTIMATED MARKOV MODEL

We deal with the restoration of grey level images corrupted by additive independent Gaussian noise. The noise variance is unknown but a signal-to-noise ratio must be given. We start with the technique proposed by S. Geman and D. Geman. The spatial interaction is described by a Markov Random Field whose energy is chosen in order to preserve the edges. This energy is parameterised. An estimation method of the parameters is proposed that is performed before the restoration. The procedure is illustrated on simulated images and nuclear magnetic resonance images.