Study On The Speckle Reduction For Polarimetric SAR Image Based On The Similarity Of The Scattering

In recent years,the remote sensing technology of the Synthetic Aperture Radar (SAR) having been developed rapidly. It has many imaging advantages, such as the mode of multi-band, multi-polarization, high-resolution and can provide rich image information for earth observing. The speckle in SAR image has seriously affected the conduct of the automatic interpretation technology. Therefore, speckle reduction of SAR image has become one of the widely concerned in the domain of SAR image processing.After several decades of research, many speckle reduction methods have been proposed, such as the Lee filter, the refined Lee filter, the Gamma MAP filter, and so on. The multi-polarization SAR can provide richer information of the objects. With the development of the multi-polarization technology, a number of speckle reduction methods for multi-polarization SAR images have been proposed, such as the polarimetric whitening filter (PWF), the polarimetric Lee filter, and so on. The common problem of the speckle reduction methods lies in their inefficiencies in details retaining and the polarimetric information preserving in the current study of the speckle reduction, which is un-conducive to the interpretation of SAR image. This paper conducts the in-depth research in the speckle reduction. The main contents are as follows:Firstly, review and summarize the SAR imaging, theoretical background, the process of electromagnetic scattering, the representation of polarimetric SAR data. Further analyze the statistical properties of the SAR data, the data model, the formation process of the speckle and its basic features. On the basis of that, study the typical speckle reduction algorithms for uni-polarimetric and multi-polarimetric SAR. And implement the typical algorithms, including the polarimetric mean filter, the PWF filter, the polarimetric Lee filter, and so on.Secondly, propose a new method to measure the similarity of the scattering characteristics. From the redundant information appearing in SAR images, this method analyze that the similar image blocks have a certain similarity of the scattering characteristics. Measure the similarity of the scattering characteristics by exploiting the Wishart distribution of polarimetric SAR data in the framework of the bayesian optimal decision theory. Thirdly, propose a new method to reduce the speckle in the polarization domain. This method introduce the similarity of the scattering characteristics into the speckle reduction of the polarimetric SAR image and draw on the thought of the weighted average filtering. In this method, the similar scattering characteristics in the neighborhood are used to restore the scattering properties of the data to be processed. If the similarity is greater, then the contribution is greater in the restoring process. At the same time, propose the calculation method of the weight by combining the new measurement of the similarity of the scattering characteristics.Fourthly, propose a speckle reduction method with edge and details retained based on the above algorithm. Join the searching of the maximum homogeneous region into the process before the speckle reduction. The searching process is to select the size and shape of the filtering window adaptively.Finally,analyze many algorithms and compare their effects on the speckle reduction through considerable amounts of true SAR data experiments. The results show that the new method is effective in improving the performance of despeckling and the preserving for details. Besides, it has an advantage in preserving polarimetric properties.