| 1. |
- Wang, Wei, et al.
(författare)
-
Superpixel-Based Segmentation of Polarimetric SAR Images through Two-Stage Merging
-
Annan publikation (övrigt vetenskapligt/konstnärligt)abstract
- This paper presents a superpixel-based segmentation method for multilook polarimetric SAR (PolSAR) data. By exploring the PolSAR statistics, a two-stage merging strategy is proposed to improve the segmentation efficiency and accuracy. First, based on the initial superpixel partition, the Wishart-merging stage (WMS) simultaneously merges the regions in homogeneous areas, which can be modelled by the Wishart distribution. The Wishart energy loss together with the edge penalty is utilized so that the merged superpixels are from the same land cover and without ambiguity. In the second stage, the doubly flexible two-parameter KummerU distribution is applied to better characterize the resultant regions from the WMS, which are usually located in heterogeneous areas. This KummerU-merging stage (KUMS) iteratively merges the adjacent regions based on the KummerU energy loss. In addition, the edge penalty and the proposed homogeneity penalty are also adopted to guide the merging procedure, and prevent merging the regions from distinct land covers. Compared with the classical iterative merging methods, the two-stage merging strategy can improve the efficiency based on the WMS, and increase the segmentation accuracy through the KUMS. Two experimental PolSAR datasets acquired by ESAR and EMISAR system are used to demonstrate the effectiveness of the proposed segmentation method.
|
|
