An Unsupervised Segmentation Using the Data Log-Likelihood for Fully Polarimetric SAR Data Analysis

In this paper, an unsupervised segmentation method is proposed for fully polarimetric SAR data. The SPAN is used to separate the data into 3 parts. Then, in each of the 3 parts, we perform the H/(R)/A initialization, the merging algorithm and the estimation of the number of clusters to achieve an unsupervised segmentation. We try to keep the definition of the SPAN as long as possible during the procedure. The experimental results show that the proposed segmentation algorithm is very fast, and the performance of the segmentation still need further investigation.