SAR Image Denoising Algorithm Based On Non-local Similarity Model In Multiscale Geometric Transform

Based on the principle of synthetic aperture,Synthetic aperture radar(Synthetic aperture,radar,SAR)achieves high resolution microwave imaging,which has the characteristics of all day long and all-weather,high resolution,large width and so on.At first,it is mainly used for airborne and space borne platform.With the development of technology,there are various applications in military and civil fields,such as the missile borne SAR,UAV,ground SAR,near space platform SAR,hand held devices and other synthetic aperture radar platform.As the SAR image is formed by superposition of coherent signal receiving terrain scattering echo,it is vulnerable to interference of coherent noise,which seriously affects the subsequent processing of SAR images,such as target extraction and target classification.etc.Therefore,it is of great significance to study the coherent noise suppression algorithms of SAR images.Since the multi-scale geometric transform not only has good time-frequency characteristics,but can better represent the image information,it has been widely used in SAR image de-noising.In addition,due to the redundancy of the image itself,there is a similarity between the pixels of the image.Many scholars apply the non local similarity of image to SAR image de-noising and get good results.On the basis of previous studies,this paper makes an in-depth study from the following two aspects.On one hand,we improve the properties of the existing multiscale geometric transformations and create new multiscale geometric transformations with better properties.On the other hand,based on the analysis of the advantages of multiscale geometric transform,the coherent noise suppression of SAR images is presented.The main research is as follows:1.SAR image de-noising algorithm based on complex wavelet transform.In view of the disadvantages of directional selectivity and invariance of the shear wave transform,this paper presents a new method of shear complex wavelet transform based on the characteristics of shear wave and dual tree complex wavelet transform.As the non sampled complex wavelet transform,the proposed method preserves the low frequency sub-bands while has better sparsity.Then,this paper proposes a new SAR image de-noising algorithm based on the combination of K-SVD algorithm,soft threshold denoising algorithm and Lee filter.Compared with other algorithms,this algorithm considers the SAR high-frequency part of the image de-noising and low-frequency part of SAR image noise suppression.The simulation results show that the proposed de-noising algorithm is effective and reliable.2.SAR Image De-Noising based on GNL-Means with Optimized Pixel-Wise Weighting in Non-Subsample Shearlet Domain.In order to overcome the problems of the artificial texture transform domain algorithm and the non-local means algorithm weight selection,preserve the image texture information as much as possible and suppress the coherent noise better,combining Non-subsampled shearlet transform and generalized non-local mean algorithm,we proposes a generalized non-local mean SAR image de-noising algorithm based on Non-subsampled shear wave domain under weight optimization.In the process of weight optimization,this paper makes use of the redundant information in overlapping blocks,and uses the two programming to solve the optimal weights in the mean square error.The experimental results show that the proposed algorithm has better de-noising effect.