SAR Image Denosing Based On Nonlocal Similarity And Low Rank Matrix Approximation

Synthetic aperture radar images can provide useful information for many applications,such as military reconnaissance and identification,remote sensing mapping,landmine detection,etc..But due to the large number of randomly distributed scatterers are the reflection of the imaging sensor,radar echo coherent superposition,which inevitably produce speckle in the image multiplicative noise,segmentation,image feature extraction for subsequent SAR brought great inconvenience.Therefore,it is of great significance to further suppress speckle noise of SAR images.The main content of this paper is the low rank matrix approximation algorithm is applied to SAR image denoising,the matrix low rank approximation algorithm are studied and improved,proposes a new algorithm for image denoising is similar to measure total variation regularization weighted kernel norm minimization based on the main contents and innovations are as follows:1.Aiming at the problem that the noise reduction effect of SAR image is not obvious in the case of low number of images,a new method of joint block matching is proposed.According to the local similarity of SAR image,using non local WNNM algorithm is applied to SAR image denoising;and then through the two-dimensional discrete cosine transform,the frequency domain information as the main reference information,proposes a matching method based on frequency domain structure similarity Euclidean distance and joint block;finally through the new joint block matching method for similar blocks.The construction of the low rank matrix,WNNM algorithm using matrix low rank approximation,the de noised images are obtained by stacking reset.The simulation results show that the SAR image denoising algorithm based on joint similar block matching and WNNM(P-WNNM)can effectively enhance the detail information of the image.2.In order to solve the problem of edge blurring of P-WNNM algorithm,an image denoising model based on total variation regularization and P-WNNM(TV-PWNNM)is proposed.Because the TV-PWNNM algorithm in the denoising iteration also filters outgeometry information,useful therefore,using principal component analysis in weak texture state from a small amount of geometric information for re estimation of residual noise variance,then using the improved TV-PWNNM algorithm to estimate the residual noise variance to matrix low rank approximation in low rank model.To reach the purpose of inhibiting SAR image noise.The simulation results show that the algorithm can preserve the details and edge information of the original image,and the method is superior to the current image denoising algorithm.