Edge detection in polarimetric SAR images based on the nonsubsampled contourlet transform

摘要

Due to the presence of speckle noise, the detection of edge features from synthetic aperture radar (SAR) images is complicated. In this paper, a new method is proposed for edge detection of polarimetric SAR (Pol-SAR) images, based on the nonsubsampled contourlet transform (NSCT). Firstly, we develop the nonsubsampled contrast pyramid (CP) for polarimetric images by introducing an adaptive optimal polarimetric contrast enhancement (OPCE) model, then the obtained multiscale images are filtered by the nonsubsampled directional filter banks. The point-wise modulus maximum is evaluated across of multidirection subbands in each scale. Finally, the multiscale edge images are fused to reduce the influence of noise. The performance of the proposed method is demonstrated with Pol-SAR data acquired by Convair-580, showing great potential of this method.