A joint classification technique based on multivariate annealing segmentation and "H/A//spl alpha/_" decomposition for fully polarimetric SAR images of suburban areas

A joint classification technique is proposed for fully polarimetric SAR images of suburban areas. The technique is based on the joint use of H/A//spl alpha/_ decomposition and multivariate annealing segmentation. This allows one to exploit the information available in the noisy estimate of the anisotropy parameter, retaining the highest possible spatial resolution. This is obtained by jointly processing the anisotropy with entropy and /spl alpha/_ angle, so that it is regularized by their more stable values. The application of the proposed technique to a AIRSAR image of suburban areas shows its practical effectiveness.