基于SURE-LET和非张量积小波的遥感图像去噪  

SURE-LET and non-tensor wavelets based remote sensing image denoising

作　　者：曾武[1,2] 徐正全[1] 周龙[2] 

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079 [2]武汉工业学院电气信息工程系,湖北武汉430023

出　　处：《华中科技大学学报（自然科学版）》2012年第2期97-100,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61075015)

摘　　要：针对遥感图像中的高斯噪声,提出了基于SURE-LET和非张量积小波的去噪方法,主要包括图像在非张量积小波下的分解、各个子带在不同阈值函数下的处理以及它们最优的线性组合3个步骤.通过选择合适的非张量积小波滤波器参数,使无噪遥感图像和噪声在变换分解中得到的小波系数分离较好,去除噪声对应的小波系数时被去除的无噪图像对应的小波系数较少,从而取得更好的去噪效果.实验结果表明:此方法用于高斯噪声的遥感图像的去噪不仅速度很快,而且去噪效果优于传统基于张量积小波的SURE-LET方法.A novel method to address the Gaussian noise of remote sensing image using non-tensor wavelet and SURE-LET was presented, which mainly contained three parts, the non-tensor wavelet decomposition, coefficients shrinkage of each subbands using the threshold functions, and estimatingthe optimal combination weights of the processed subbands. The non-tensor wavelet filters could be represented in the parametric form, by using appropriate, the non-tensor wavelet coefficients of noise free remote sensing images and noise were separated better than the traditional tensor wavelet coeffi-cients. As a result, when using coefficient shrinkage technique to remove the noise, more noise free image coefficients could be reserved. Consequently, better denoising performance could be obtained. Experimental results show that by combining non-tensor wavelet and SURE-LET, the denoising pro- cedure is very fast and denoising performance in sense of PSNR is prior to tensor wavelet and SURE- LET.

关 键 词：遥感图像 高斯噪声 图像去噪 Stein无偏风险估计 非张量积小波 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]