An Unsupervised Segmentation Using SPAN/H/γ/A initialization for Fully Polarimetric SAR Data Analysis

In this paper, an unsupervised segmentation method is proposed for fully polarimetric SAR data. We use the parameter SPAN combined with the H/γ/A to perform the initialization, and the Wishart test statistic is used reduce the number of clusters. The output number of clusters is determined by the data log-likelihood algorithm. We try to keep the definition of SPAN as long as possible during the segmentation procedure. The experimental results show that the proposed segmentation algorithm is very fast, but the performance of the segmentation still need further investigation.