Prior model :

First order neighbourhood

Modified Geman & Yang sampler, or Gibbs or Metropolis samplers

=0.3, =10	=1, =15	=1, =5

Posterior distribution :

Data dependence : observed image Y

High order neighbourhood (depends on the size of H)

Only the modified Geman & Yang sampler is possible
        (Gibbs or Metropolis are too time consuming because of the high order neighbourhood)

=0.5, =10	=0.5, =1	=10, =10

Observation image Y :