Research On SAR Image Preprocessing For Feature Similarity Of Same Scene Optical Image

With the successful launching of satellite-borne SAR in our country,the data resource of satelliteborne SAR image is increasingly rich.The satellite-borne optical image acquisition is effected by the confines of night and day and the weather condition of the atmosphere,and SAR image acquisition can effectively remedy the defect due to the capacity of all-weather and all-time.But the interpretation of SAR image is more complicated than optical image,and for human vision,the intuitive sense of SAR image is under optical image.Therfore,it is urgent to transform SAR image into optical image or make them similar.The special imaging principle of SAR would cause serious speckle noise,high dynamic range image,and geometric distortion of SAR image.Aiming at SAR image preprocessing technology,the suppression of SAR image speckle based on sparse representation and transform domain,the enhancement of high dynamic range image and feature similarity of same scene optical image are discussed in this paper.The main research work and results of this paper are shown as follow:(1)A SAR despeckling algorithm where the sparse structure meets Gaussian Scale Mixture(GSM)model is proposed.The mathematic model derivation of the algorithm is based on Bayes principle and the statistical property of speckle noise.The probability is used to measure the weight in the block matching process;by utilizing the structural similarities of the image blocks,the homogeneous region and heterogeneous area are efficiently classified,and a better mean estimated value of the block is get.The model is solved via iteration and regularization,and a local optimized solution of the denoised image matrix is obtained.Experimental results show that the algorithm has a competitive performance in terms of despeckling and the protection of local structural characteristics and textural features.(2)A non-subsampled Contourlet transform(NSCT)domain filtering algorithm based on adaptive shrinkage and GSM sparse representation is proposed for SAR image despeckling.The statistical property of speckle in high frequency subbands is studied in this part.Therefore,the despeckling process of high frequency subbands in NSCT domain is translated into recovering the sparsity of the high frequency coefficients.Firstly,the priori ratio and posterior ratio are used to get the mask estimation of the high frequency coefficients,this process can achieve adaptive shrinkage,which would remove a part of speckle and reduce the noise impact of the dictionary training in sparse coding step.And then,the GSM model is used in the process of high frequency coefficients sparse representation.Experimental results show that the proposed algorithm achieves outstanding denoising result and edge preservation ability.(3)An optimal contrast-tone mapping method is used in image enhancement for high dynamic range SAR image.The tone continuity is added to the contrast gain model as a constraint condition,and it can avoid tone distortion when the contrast gain is maximum.A linear programming approach is used in the solution of the optimal model,and it is simply effective.Experimental results show that it has great effect in enhancement of SAR image which has dark background information and luminous object information,and improves the image contrast effectively.(4)The processing method of SAR data based on ENVI is studied.The SLC SAR data is processed by multi-look,filtering,geographic coding and calibrating on ENVI by using SARscape.And the similarity between the processed SAR image and the same scene optimal image is assessed by the index which based on image structure information.Experimental results show that the similarity between the processed SAR image and the same scene optimal image is improved.