Polarimetric SAR Despeckling Research Based On Local And Nonlocal Strategy

Polarimetric synthetic aperture radar( POLSAR) is a kind of multiparameters, multichannel microwave imaging radar and has certain transmissivity to ground objects. It can provide the target scattering echo under different state and multidimensional information, and is the new technology of remote sensing, which is widely applied in the field of military, defense, environment, agriculture and so on. But because of the principle of the coherent imaging, the information get b y polarimetric SAR has certain deviation with the actual information. Speckle noise, which affects the interpretation and the subsequent application, appears in the PO LSAR image. So the primary problem in the application of polarimetric SAR data is the speckle noise suppression. In this paper the statistical distribution and coherent speckle noise model of SAR and polarimetric SAR data are introduced and analyzed. Three kinds of speckle noise suppression algorithm are put forward and the main work is as follows:1. In consideration of the existing bilateral filtering of polarization SAR and the edge detection method based on relative total variation, a kind of speckle noise suppression method, which detects the homogeneous area and edge details area and sets the different brightness values in different regions, is proposed for polarimetric SAR data. This method is simple and implement easy. The proposed method can deal with polarimetric covariance matrix C and polarimetric coherent matrix T directly. Experiments show that this method not only can suppress coherent speckle but also maintain the detail information of the edge.2. Combined with bilateral filter and non- local means filter, a Non Local-bilateral polarimetric SAR despeckle method is proposed, which selected the similar patch in nonlocal area, adopt bilateral filtering over those similar patch and nonlocal average the results by bilateral filter. This method considers the structure information as well as the single pixel information of image. The proposed method combines the advantages of bilateral filter and non- local means filter. Experiments show that this method can achieve good results not only in homogeneous area but also on edge pixel.3. Extending the additive white Gauss noise SVD framework to SAR dspeckle, in consideration of the statistical distribution and the noise model of SAR data, instead of original Euclidean distance, SAR similarity distance is used to build SVD sample matrix. According to SAR image speckle characteristics, a suitable SVD threshold parameter is derived. In addition, considering the characteristics of polarimetric SAR data, expanding the modified SVD framework to POLSAR despeckle. In order to maintain polarization information, the similar image patch and SVD threshold parameter is obtained over Span data. The experiments show that SAR data and POLSAR data dspeckle has achieved very good results, the greater advantage of the effectiveness of single-polarimetric SAR data processing.Our work is supported by national natural science foundation of C hina(No.61173092), new century excellent talents to support plan(No.66ZY110) and shaanxi province science and technology research and development projects(No.2013KJXX-64).