Speckle Filtering Method For PolSAR Data Based On Matrix Log Cumulant And Non-local Approach

Polarimetric Synthetic Apeture Radar(Pol SAR)is a multi-channel imaging radar system.Since targets with different physical properties have different back scattering features when being detected by different polarimetric microwaves,Pol SAR system measures back scattering features of the targets in each resolution cell through transmitting,receiving and processing the radar signals in several separated polarimetric channels.Therefore,compared with the conventional SAR,Pol SAR system can provide more information of the target and bring more possibilities for the successive Radar image processing in a more convenient way.In recent years,with the development of the hardware performance,the emphasis of Pol SAR has gradually moved from theory to engineer application.However,Pol SAR image speckle noise is more complicated than that of the conventional SAR.Besides,their mathematical models are also different.Therefore,Pol SAR image speckle noise suppression has become one of the hot issues in this technical field.This thesis has conducted an in-depth research in this field,which applied the new research findings of mathematical technique with regard to matrix log-cumulant for Pol SAR data to Pol SAR image speckle noise suppression technique.A theoretical method of calculating the Pol SAR sub-look data correlation based on matrix log-cumulant is proposed in this thesis.In recent years,complex matrix log-cumulant based mathematical tools have been introduced into the statistical research of complex scattering matrix of Pol SAR data,and a series of achievements have been achieved.However,these mathematical tools are mostly used as the statistical analysis tools or applied in some research fields of image quality measurement,but are still not applied in the key technique of Pol SAR image speckle noise suppression.This thesis proposed a new theoretical method of calculating the Pol SAR sub-look data correlation based on matrix log-cumulant,providing a new thought for applying this theory in the techniques of Pol SAR image speckle noise suppression and classification.The main achievements of this research are as follows: 1.A Pol SAR image noise suppression algorithm based on non-local algorithm with filtering coefficient calculated from matrix log-cumulant of sub-look Pol SAR data is proposed.In this algorithm,the filtering coefficient can be calculated from the theoretical method based on the correlation of sub-look Pol SAR data.And with the non-local based homogeneous region selection,the new algorithm is introduced.This new algorithm is tested and verified through experiments,and compared with previous ones.Meanwhile,the aforementioned theoretical method of calculating the Pol SAR sub-look data correlation is verified by the experimental results.2.A non-local Pol SAR image noise suppression algorithm with matrix log-cumulant based homogeneous region selection and weighting coefficient calculation,is proposed.When measuring the correlation of two complex matrix sets by using matrix log-cumulant,as many elements of complex matrix sets as possible are needed.In the previous research,although the filter coefficient is successfully calculated based on the correlation of sub-look Pol SAR data,there are still some shortcomings when selecting the homogeneous region.Referring to this problem,this thesis selected more multi-look data points using edge allied window to extend the sub-look data set.Thus,a more accurate method of calculating the correlation coefficent and selecting the homogeneous region is derived.The effectiveness of this algorithm is verified through experiments,and the results are compared with other algorithms.