Using wavelet-domain hidden Markov models (HMMs) can efficiently improve the performance of SAR image information extraction

where the key problem is how to estimate the hidden state under the influence of multiplicative speckle noise

which is not solved effectually yet.Making use of the principle of turbo iterative decoding in information theory and aiming at SAR image

the paper proposed a new method of turbo iterative estimation for the hidden state

in which the hidden state is estimated alternatively with two different restriction conditions in two uncorrelated sub-spaces

and the posterior probability

as an kind of extrinsic information

is exchanged between the two sub-spaces.The experimental results illustrate that the method has a rather impressive performance and very fast convergence rate.