A novel unsupervised change detection for synthetic aperture radar (SAR) images using robust semi-non-negative matrix factorisation (NMF) with total variation (TV) is proposed. In order to effectively handle the noise… Click to show full abstract
A novel unsupervised change detection for synthetic aperture radar (SAR) images using robust semi-non-negative matrix factorisation (NMF) with total variation (TV) is proposed. In order to effectively handle the noise and outliers as well as consider spatial contextual information, correntropy induced metric and TV regularisation are simultaneously integrated into semi-NMF to build a robust semi-NMF with TV model. Subsequently, the update solutions of the model are optimised by the Karush–Kuhn–Tucker condition and fast gradient projection algorithm. The experimental results demonstrate the effectiveness of the proposed method both visually and quantitatively.
               
Click one of the above tabs to view related content.