Multitemporal SAR images change detection based on joint sparse representation of pair dictionaries

Wei Li, Jiayu Chen, Pei Yang, Hong Sun

Abstract: In this paper, we present a novel automatic and unsupervised change detection method specifically oriented to the analysis of multitemporal synthetic aperture radar (SAR) images. This object-based method takes full advantage of both the dictionary learning method and the sparse representation theory for very high resolution (VHR) SAR images. Under our scheme, a pair of local dictionaries is obtained by the K-SVD dictionary learning method in each region segmented from original SAR images, and then the change detection is executed by the joint sparse representation approach through a comparison of the coefficient matrixes of the regional SAR data representing on a dictionary with their mix-norms. This method is applied on two groups of TerraSAR-X data, and the results show that it is robust and outperforms the previous methods related.