Beta‐divergence based two‐phase segmentation model for synthetic aperture radar images. Issue 20 (9th September 2016)

• Record Type: Journal Article
• Authors: Woo, H.
• Is Part Of: Electronics letters. Volume 52:Issue 20(2016)
• Page Start: 1721
• Page End: 1723
• Publication Date: 2016-09-09
• Subjects: synthetic aperture radar -- radar imaging -- image segmentation; beta‐divergence based two‐phase segmentation model -- synthetic aperture radar image -- beta‐CCV model -- high dynamic range SAR image -- convex‐relaxed Chan‐Vese model -- convex optimisation; Electronics -- Periodicals; 621.381
• DOI: 10.1049/el.2016.2798
• Languages: English
• ISSNs: 0013-5194

Abstract

Abstract : A general two‐phase segmentation model (beta‐CCV) for synthetic aperture radar (SAR) images is introduced. The proposed model is based on the beta‐divergence similarity measure and the convex‐relaxed Chan–Vese model. The main advantage of beta‐divergence with non‐positive value of beta is that it is an expedient similarity measure for high dynamic range SAR images with strong scatterers. In addition, due to the global optimum property of beta‐divergence, naturally merge the beta‐divergence similarity measure into the convex‐relaxed Chan–Vese model. Hence, we can find solutions of beta‐CCV in a more stable way through convex optimisation. The proposed beta‐CCV model shows overall better performance than the state‐of‐the‐art segmentation method.