基于NSCT和自适应模糊阈值遥感图像去噪算法  

Remote Sensing Image Denoising Algorithm Based on NSCT and Adaptive Fuzzy Threshold

作　　者：黄涛[1,2] 薛丰昌[3] 钱洪亮[3] 周明 

机构地区：[1]上海市嘉定区气象局,上海201800 [2]南京信息工程大学计算机与软件学院,江苏南京210044 [3]南京信息工程大学遥感学院,江苏南京210044

出　　处：《计算机技术与发展》2016年第1期65-69,共5页Computer Technology and Development

基　　金：国家级城市气象科学研究基金(IUMKY&UMRF 201103);江苏省博士后科研基金(1101024B);南京信息工程大学科研基金(S8111133001)

摘　　要：针对遥感图像去噪过程中硬阈值去噪的伪吉布斯(Gibbs)现象带来的视觉失真、软阈值去噪对图像细节造成的"过扼杀"现象以及峰值信噪比(PSNR)较低等问题,提出一种基于非下采样Contourlet变换(Non-Subsampled Contourlet Transform,NSCT)和自适应模糊阈值的遥感图像去噪算法。该算法通过对含噪图像进行NSCT变换,从而得到不同尺度与方向上的NSCT域系数,然后结合噪声的分布特点,基于贝叶斯萎缩法,对Bayes Shrink阈值进行了改进,并结合模糊理论,构造模糊阈值函数,实现对变化域内的系数进行处理,最后将处理后的系数进行NSCT反变换,以得到去噪后的图像。实验结果表明,该方法不仅很好地克服了硬阈值去噪的伪吉布斯现象带来的视觉失真、软阈值去噪对图像细节造成的"过扼杀"现象,而且能有效地提高峰值信噪比,保持遥感图像丰富的纹理信息。In order to overcome the visual distortion caused by Gibbs phenomenon of hard thresholding,the ＂ over strangle＂ phenomenon caused by processing image detail of soft thresholding,the lower PSNR and other issues in the process of image denoising,an algorithm of denoising for remote sensing images based on NSCT and adaptive fuzzy threshold was proposed. First,transform the noisy image by NSCT to get the NSCT domain coefficients of different scales and different directions. Then improve the Bayes Shrink threshold with the characteristics of the noise distribution based on Bayesian shrinkage method,constructing a fuzzy threshold function with fuzzy theory to process the coefficients within the changing domain. Last,inversely transform the processed coefficients by NSCT to get the denoised images. Experimental results showthat the proposed algorithm can not only overcome the visual distortion caused by Gibbs phenomenon of hard thresholding,the ＂ over strangle＂ phenomenon caused by processing image detail of soft thresholding well,but also improve PSNR effectively,keeping the rich texture information of remote sensing images.

关 键 词：非下采样CONTOURLET变换 模糊阈值 遥感图像 去噪算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]