In this paper

a kind of SAR image segmentation method based on the criterion of likelihood difference function is proposed. At first

based on the distribution property of SAR image

we get the following conclusions by simulation: After merging of two regions with same distribution

the difference of likelihood is approximately independent of the regional size and regional mean value

and only depends on the look number. On the basis of this conclusion

we simplify the likelihood difference function

and the analytic form of the probability distribution function is obtained. Giving the false alarm probability

we can calculate the threshold which indicates whether the adjacent regions can be merged. Then we construct merging cost according to the likelihood difference function and region shape constrains

and make regions to be merged with order. When there are no regions which can be merged

the final segmentation result is achieved.