Research On Speckle Suppression And Feature Detection In SAR Image

A large number of grainy speckle noise caused by coherent imaging in synthetic aperture radar(SAR)images,these speckles with multiplicative noise attributes have caused serious interference to the subsequent interpretation of SAR images.Edge detection is one of the basic aspects of SAR image feature detection,and its detection is also affected by speckles,therefore,this paper focuses on the study of SAR speckle suppression and edge detection,and proposes two speckle suppression algorithms and an edge detection algorithm.First,a SAR image speckle suppression algorithm based on mean ratio and information entropy is proposed.The algorithm uses the local mean ratio of each pixel in the search window and the center pixel in the similarity window to construct similarity measurement parameters,constant false alarm rate(CFAR)characteristic of edge detection in SAR image based on mean ratio,improve NLM algorithm for image edge detection capability of low contrast region on the SAR image;Using the information entropy of the SAR image after the logarithm and Gauss smoothing to construct the attenuation factor,and using information entropy for edge sensitivity,adaptively control the weighted average strength of the NLM algorithm in the homogeneous area and the edge area,ensure that the edge details are further protected while increasing the intensity of the speckle;The introduction of non-local mean idea makes more pixels participate in the weighted average operation,and improves the performance of SAR image edge protection and speckle suppression in an all-round way.Second,a SAR image suppression algorithm based on wavelet packet is proposed.The algorithm reconstructs the attenuation factor on the basis of the above algorithm,and uses the image after the logarithm processing to decompose the wavelet packet.The calculation of information entropy by low frequency information after wavelet packet decomposition,can reduce the noise impact on the information entropy,the information entropy calculation is more accurate,then the normalized information entropy as the new attenuation parameters of the algorithm,the weighted average is carried out under the idea of non-local mean,which makes the SAR image speckle suppression and edge protection be further improved.At last,a SAR image edge detection algorithm based on anisotropic Gauss kernel is proposed.The algorithm uses an improved Gauss gamma window instead of the traditional rectangular window to calculate the edge intensity mapping and direction information,the anisotropic Gaussian kernel function is used to detect the edge of SAR image,experimental results show that the false edges detected by this algorithm are significantly reduced,and the edge-preserving performance is improved compared with other algorithms.This paper first introduces the research background,significance and the imaging principle of SAR system,then expounds the classic speckle reduction algorithm and edge detection algorithm,and then two kinds of speckle suppression algorithms and an edge detection algorithm are proposed,and verified by MATLAB simulation experiment,finally,it summarizes and looks forward to the full text.