均值滤波
中值、高斯滤波
非局部均值去噪
程序:
close all;
clear all;
clc
I=double(imread('lena.tif'));
I=I+10*randn(size(I));
tic
O1=NLmeans(I,2,5,10);
toc
imshow([I,O1],[]);
function DenoisedImg=NLmeans(I,ds,Ds,h)
%I:含噪声图像
%ds:邻域窗口半径
%Ds:搜索窗口半径
%h:高斯函数平滑参数
%DenoisedImg:去噪图像
I=double(I);
[m,n]=size(I);
DenoisedImg=zeros(m,n);
PaddedImg = padarray(I,[ds,ds],'symmetric','both');
kernel=ones(2*ds+1,2*ds+1);
kernel=kernel./((2*ds+1)*(2*ds+1));
h2=h*h;
for i=1:m
for j=1:n
i1=i+ds;
j1=j+ds;
W1=PaddedImg(i1-ds:i1+ds,j1-ds:j1+ds);%邻域窗口1
wmax=0;
average=0;
sweight=0;
%%搜索窗口
rmin = max(i1-Ds,ds+1);
rmax = min(i1+Ds,m+ds);
smin = max(j1-Ds,ds+1);
smax = min(j1+Ds,n+ds);
for r=rmin:rmax
for s=smin:smax
if(r==i1&&s==j1)
continue;
end
W2=PaddedImg(r-ds:r+ds,s-ds:s+ds);%邻域窗口2
Dist2=sum(sum(kernel.*(W1-W2).*(W1-W2)));%邻域间距离
w=exp(-Dist2/h2);
if(w>wmax)
wmax=w;
end
sweight=sweight+w;
average=average+w*PaddedImg(r,s);
end
end
average=average+wmax*PaddedImg(i1,j1);%自身取最大权值
sweight=sweight+wmax;
DenoisedImg(i,j)=average/sweight;
end
end