Research On Structure Preserving Speckle Reduction For PolSAR Image Based On Bilateral Filtering

Synthetic Aperture Radar (SAR) has unique advantages in the filed of remote sensing observation for the imaging ability of all-day, all-weather and high resolution, which has widely applied in military and civilian fields. Polarimetric SAR (PolSAR) can obtain more scattering information of targets than SAR, which provides rich data for SAR images correct interpretation. There are large amounts of speckle in PolSAR images for the characteristics of coherent imaging system, which has seriously affected human interpretation and automated sense analysis in image contents. So, speckle reduction becomes the important part of image interpretation and research. Concerning the problem of speckle reduction of PolSAR images, this dissertation analyzings the key issues of structural features extraction and preservation during the despeckling, and then introduces structural characteristics and polarimetric scattering property into reducing speckle. Based upon it, structure preserving speckle reduction of PolSAR images is proposed, which improves the performance of structure preserving.The speckle reduction for PolSAR images is researched in this dissertation, and the main works and innovation are as follows:(1) The speckle reduction for PolSAR images based on edge preserving bilateral filtering is proposed. It makes full use of the advantages of polarimetric bilateral filtering of preservation of edge and polarimetric information. Upon these advantages, the novel method introduces edge structural characteristics and surface scattering features into the speckle reduction, which furthers on keeping edge and polarimetric scattering information for PolSAR images as well as despeckling, and improves the performance of speckle reduction.(2) The speckle reduction for PolSAR images by using hybrid features similarity bilateral filtering is proposed. It utilizes the extracted structural similarity index on the Total Scattered Power image to build the weighting function of structural similarity, and the theory of deorientation is applied to PolSAR data before polarimetric decomposition and unsupervised classification, which improve the accuracy of the extracted of structural information and surface scattering information. The performance of structure preserving is further improved in the proposed algorithm.To verify the effectiveness of structure preserving speckle reduction for PolSAR images based on bilateral filtering in this dissertation, some PolSAR images acquired by the airborne system of AIRSAR and Convair are tested sufficiently and supplied by detail analysis. Compared with other polarimetric bilateral filtering, the performance of structure preserving and polarimetric scattering information is improved by using the proposed methods. The proposed algorithm would be a meaningful progress towards an accurate interpretation of PolSAR images.