基于多尺度分析的图像去噪方法研究  

Research on Image Denoising Method Based on Multi-scale Analysis

作　　者：曹芳菊 CAO Fangju(Hebei Vocational University of Technology and Engineering,Xingtai Hebei 054001)

机构地区：[1]河北科技工程职业技术大学,河北邢台054001

出　　处：《软件》2021年第8期7-10,共4页Software

基　　金：邢台市重点研发计划自筹项目,基于Matlab的医学图像去噪方法研究(2020ZC039)。

摘　　要：本文研究了两种新的基于多尺度分析的阈值去噪方法,分别是基于Curvelet变换的局部Winner滤波去噪和基于非下采样Contourlet变换的自适应阈值去噪,同时在估计子带系数方差的时候,设计了弧型窗口来完成,更符合Curvelet变换和Contourlet变换各向异性的特征,用文中方法进行图像去噪实验,提高PSNR值的同时,视觉效果也比较好,保留了图像本身的一些细节特征。In this paper two new threshold denoising methods based on multi-scale analysis are studied,which are local Winner filter denoising based on Curvelet transform and adaptive threshold denoising based on non down sampling Contourlet transform,at the same time,are windows are designed to estimate the variance of subband coefficients,which is more in line with the anisotropic characteristics of Curvelet transform and Contourlet transform,the image denoising experiment is carried out with the method in this paper,while improving the PSNR value,the visual effect is also better,and some detailed features of the image itself are retained.

关 键 词：CURVELET变换 非下采样Contourlet变换(NSCT) 子带自适应阈值 弧形窗 图像去噪 

分 类 号：TN957.52[电子电信—信号与信息处理]