Despeckling And Segmentation Methods For Synthetic Aperture Radar Images

As a kind of microwave remote sensing technology,synthetic aperture radar(SAR)has the unique advantages of high resolution,strong penetrability,all-day and allweather work,and has been widely used in military and agricultural fields.The coherent imaging principle of SAR system leads to a lot of speckle noises with multiplicative characteristics in SAR images.The non-uniformity and complexity of the speckle noise seriously hinder the automatic interpretation of the SAR image in the later stage,for example,affecting the accuracy of the target segmentation,which increases the difficulty for subsequent target recognition and tracking applications.Therefore,it is of great significance to carry out research on SAR image speckle suppression and segmentation methods.This paper focuses on the speckle suppression of SAR images,and also preliminarily studies SAR image segmentation.The specific research work is as follows:By analyzing and comparing the speckle suppression performance of classical spatial filtering,a speckle suppression method based on improved Frost filtering is proposed.This method uses the classic Lee filter coefficient or Kuan filter coefficient to control the filtering intensity of Frost,and realizes the Frost weighted filtering of SAR images with different parameters in different scenes,avoiding over filtering or less filtering problems.Compared with classical airfield suppression methods such as classical Lee filtering,the experimental results show that the method has obvious advantages in visual effects and parameter indicators.By improving the classical non-local means(NLM)method,a non-local means speckle suppression method based on CV is proposed.The similarity measurement of the method uses Gaussian weighted Euclidean distance of the image after logarithmic transformation and Gaussian smoothing.The adaptive attenuation factor is composed of the reciprocal of the CV of original images.Finally,the new weights,which are composed of the new similarity measurement parameters and adaptive attenuation factors,perform non-local weighted filtering on the original SAR images.Compared with several speckle suppression methods in recent years,the experiments show that this method can improve the speckle suppression performance of homogeneous areas with high gray level while maintaining the edge.The statistical characteristics of SAR images are added to the non-local means filtering process,and a non-local means speckle suppression method with adaptive adjustment of filtering intensity is proposed.The method is based on the estimation model of Lee filter or Kuan filter,and performs non-local weighted filtering according to the local statistical information of the SAR image.The weight of the non-local weighted filtering is composed of a similarity parameter based on mean ratio(MR)and attenuation factor based on CV.Compared with the latest several speckle suppression method,the experimental results show that the method have a great improvement in both visual effects and parameters.In order to overcome the influence of speckle noise on SAR images segmentation,an active contour model(ACM)segmentation method is proposed.This method introduces the diffusion equation and diffusion coefficient based on partial differential theory into the level set ACM,which reduces the intensity of the speckle noise and more accurately segments the target region in the SAR images.Compared with several segmentation methods in recent years,the proposed method improves the segmentation performance of SAR images on the edge location accuracy index.