SAR Image Denoising And Edge Detection Based On Directionlet Transform

Synthetic aperture radar (SAR) is a variety of weather conditions can be obtained the high-resolution images radar. It is coherent imaging mechanism will inevitably produce speckle noise, so the image can not effectively reflect the target scattering characteristics of object, affecting the quality of SAR images, while not conducive to the computer's automatic analysis and understanding. Therefore, the suppression SAR image speckle noise, is an important issue of SAR imaging applications. The edge is also an important feature of the image and its detection results directly affect the SAR image recognition and analysis.In this paper, multiscale geometric analysis tool for the research, applying the latest research tool, according to its characteristics be applied to SAR image processing. Paper mainly for the following three areas:1. Due to the property of SAR images speckles is multiplicative noise, in this paper, a novel approach to SAR image speckle reduction is presented, which is based on Directionlet domain. Add Cartoon + Texture model for SAR image decomposition, and using Gaussian mixture model fitted to the Directionlet decomposition coefficients, and finally Bayes minimum mean-square error estimates for denoising, by comparing the results show that this method the feasibility and superiority.2. The improved Directionlet transform combined with the stationary wavelet is proposed nonsubsampled anisotropic wavelet transform denoising method, the use of decomposition of the transformation of non-sampling make get a better correlation between coefficients, thereby using neighborhood contraction of de-noising processing, experiments show that this approach has been good de-noising effect.3. An effective edge detection algorithm was presented in this paper, which was combined nonsubsampled anisotropic wavelet transform denoising SAR image with ratio improved operator .This denoising method use the SAR image denoising processing in 2, thus accurate positioning the edge of the original image information, and then using of the improved ratio operators on the edge of the information received the extraction of edge pixels. Experimental results validate the feasibility and effectiveness of this method.