Edge detection for polarimetric SAR images using span-driven adaptive filter

In this paper, a span-driven adaptive (SDA) filter is utilized for detecting the edges in polarimetric synthetic aperture radar (PolSAR) images. The SDA filter has variable sizes and flexible shapes, and it is constructed by adaptively selecting the independent and identically distributed samples. Therefore, it can overcome the limitation of classic fixing-shape filters, and obtain more accurate estimation of local areas. Since the complex Wishart distribution is often not met, particularly in heterogeneous urban areas, we adopt the spherically invariant random vector (SIRV) product model to obtain refined and reliable edge detection results. Experimental results on both simulated and real PolSAR datasets demonstrate the effectiveness of the proposed method and its superiority over the traditional edge detectors.