空域相关滤波SSNF算法程序.docVIP

%%%%%%%%%%%%%%%%%%%%%%%空域相关滤波%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%% %参考文献： %Wavelet transform domain filters:a spatially selective noise filtration technique %不能选正交小波，文献中提出用二次样条小波（quadratic-spline）.此处用双正交小波：bior 1.5 %优点： % 1.对于信噪比高的信号滤波效果好； % 2.对于边沿的保护强过阈值滤波，不会产生阈值滤波情况下的过于平滑与Gibbs现象。 %缺点： % 1.由于对边沿信号没做任何处理，所以边沿可能会有脉冲噪声保留下来； % 2.计算相关系数中，如果计算出来的小波系数点位置偏差大，则相关系数计算受影响； % 3.需要迭代运算，迭代的噪声能量阈值选取很重要，这里以开始段无信号处估计噪声； % 4.需要迭代运算，所以运算量比阈值法大； % 5.受分解层次影响，在大尺度上小波系数点位置偏差更大，相关系数计算不准确。 %需要具体调整的地方： % 1.分解的尺度； % 2.选定用什么信号作为噪声的估计； % 3.设定停止迭代的噪声能量阈值参数cc。 %%%%%%%%%%%%%%%%%%%%%%%空域相关滤波%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% close all; clc; clear; snr=5; init=2055615866; [xref,x]=wnoise(1,10,snr,init); signal=x; points=1024; level=5; wf=bior 1.5; %sym8,bior 1.5 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % %阈值消噪： S_thr=wden(signal,rigrsure,s,mln,5,sym8); % % rigrsure;heursure;sqtwolog; minimaxi % %one;sln;mln % subplot(211);plot(signal); % title(阈值滤波); % subplot(212);plot(S_thr); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %进行二进制小波变换（离散平稳小波变换），并给出各级波形： [Lo_D,Hi_D,Lo_R,Hi_R]=wfilters(wf); [swa,swd] = swt(signal,level,Lo_D,Hi_D); % figure; % subplot(level,1,1); plot(real(signal)); grid on;axis tight; % for i=1:level % subplot(level+1,2,2*(i)+1); % plot(swa(i,:)); axis tight;grid on;xlabel(time); % ylabel(strcat(a ,num2str(i))); % subplot(level+1,2,2*(i)+2); % plot(swd(i,:)); axis tight;grid on; % ylabel(strcat(d ,num2str(i))); % end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %小波系数的处理： Swd_n=swd; swd_org=swd; mask_n=zeros(size(Swd_n)); %先把系数处理矩阵设置为全0。 for j=1:(level-1) %在1:(level-1)分解层次上对高频系数处理，最后一层无法求相关系数，所以不作处理。 Noise_d1=swd_org(j,:); Noise_d1=Noise_d1(1:80); Noise_var=var(Noise_d1); %以信号的前80个只含有噪声的点估计噪声在各层的方差