Research On Decoupling Method Of SAR Image Denoising And Deblurring

Because of the ability of interferometry and multipolarization measurement and the advantages of all-weather,all-weather and strong penetration,synthetic aperture radar(SAR)has the advantages that optical remote sensing can not compare.However,the quality of SAR image will be seriously affected by the speckle noise and the blur factors of imaging system.It is the most challenging task to get the ideal SAR image which is close to complete non degradation by the restoration algorithm.In this thesis,the decoupled method of image denoising and deblurring dual spatial filtering is used.By improving the non local means(NLM)denoising algorithm and the point spread function estimation algorithm of imaging system,the high-quality restoration of SAR image with noise and fuzzy is realized.The main work is as follows:Firstly,the decoupled scheme of denoising and deblurring dual spatial filtering is discussed.The simulation model is established by using relatively clear SAR image.The performance of different decoupling methods is compared from PSNR and computational complexity.It is found that the performance of the decoupling method based on Wiener filter is the best,PSNR can be greatly improved,and the computational complexity of the algorithm can meet the real-time requirements.Therefore,the decoupled algorithm framework of SAR image denoising and deblurring reduction is established.In order to improve the performance of decoupling algorithm and its denoising ability,the nonlocal mean algorithm is improved.The traditional neighborhood collection strategy is changed into adaptive neighborhood collection strategy;the Euclidean distance is replaced by the mean ratio which is more suitable for SAR image noise model as similarity measurement criterion;and the fixed attenuation factor is changed into adaptive attenuation factor.The experimental results show that the denoising effect and edge preserving ability of the algorithm are better than the original nonlocal mean algorithm and SARBM3 D algorithm which are widely used at present.In order to improve the decouple performance and the ability of deblurring,a blind estimation algorithm based on the blur kernel of SAR image is discussed.It is found that Laplace distribution model is more suitable for the blur kernel modeling of SAR image.Therefore,in the logarithmic domain,the ratio of the radial power spectrum of different scales of the gradient SAR image is used as the data fidelity term to estimate the parameters of the blur kernel based on the Laplace distribution.A series of frequency-domain polynomials are used to approximate the reciprocal of the fuzzy kernel,and the dual representation of the time-domain is calculated to obtain the fuzzy kernel.Simulation results show that the algorithm can achieve better restoration effect with little time-consuming.Finally,the denoising and deblurring algorithm is used in the dual spatial filtering decoupling framework,and the final de-noising and de blurring decoupling method of SAR image is formed by optimizing the parameters.The test results show that the decoupled method can effectively remove the noise and blur in the image and improve the definition of the image.