均值、中值、高斯、维纳、小波、NL-means的图像去噪的matlab程序

1.均值、中值、高斯、维纳图像去噪

clc
clear
g = rgb2gray(imread('chuzao.jpg'));
zyy = imread('y.jpg');
g1 = imfilter(g, fspecial('average'));%均值滤波
g2 = medfilt2(g, [3 3]); %3*3中值滤波
h_gaosi1 = fspecial('gaussian',3,1);
g3 = imfilter(g, h_gaosi1); %高斯平滑滤波
g4 = wiener2(g,[5 5]);  %5*5维纳滤波
subplot(3, 2, 1), imshow(zyy), title('原图');
subplot(3, 2, 2), imshow(g), title('待处理噪声');
subplot(3, 2, 3), imshow(g1), title('均值滤波');
subplot(3, 2, 4), imshow(g2), title('中值滤波');
subplot(3, 2, 5), imshow(g3), title('高斯平滑滤波');
subplot(3, 2, 6), imshow(g4), title('维纳滤波');
%下面是计算PSNR和SSIM的程序口令
psnr(zyy,g)
psnr(zyy,g1)
psnr(zyy,g2)
psnr(zyy,g3)
psnr(zyy,g4)
ssim(zyy,g1)
ssim(zyy,g2)
ssim(zyy,g3)
ssim(zyy,g4)

2.小波重构除噪

clc
clear; 
zyy = imread('y.jpg'); %读取原图像
subplot(221);
imshow(zyy); 
title('原图');   
X=imread('chuzao.jpg');            
X=rgb2gray(X);
subplot(222);          
imshow(X);             
title('待处理噪声');                        
X=double(X);
%用小波函数coif2对图像X进行2层
% 分解
[c,l]=wavedec2(X,2,'coif2'); 
% 设置尺度向量
n=[1,2];                  
% 设置阈值向量 , 对高频小波系数进行阈值处理
p=[10.28,24.08]; 
nc=wthcoef2('h',c,l,n,p,'s');
% 图像的二维小波重构
X1=waverec2(nc,l,'coif2');   
subplot(223);              
imshow(uint8(X1));                
%colormap(map);            
title(' 小波第一次消噪后的图像 '); 
%再次对高频小波系数进行阈值处理
mc=wthcoef2('v',nc,l,n,p,'s');
% 图像的二维小波重构
X2=waverec2(mc,l,'coif2');  
subplot(224);             
imshow(uint8(X2));               
title(' 小波重构第二次消噪后的图像 '); 

 

 

3.NL-means除噪

clc
clear
g = rgb2gray(imread('chuzao.jpg')); %读取待去噪图
zyy = imread('y.jpg');%读取原图
g1 = double(g);
%进行NL-means除噪 
[m,n]=size(g1);
ds=2;% block size for calculate weight
Ds=5;% search block
h=10;% decay factor
offset=ds+Ds;
PaddedImg = padarray(g1,[ds+Ds,ds+Ds],'symmetric','both');% 扩展图像，便于处理
%距离加权核

%非均值核
[x,y]=meshgrid(-ds:ds,-ds:ds);
kernel=1./(x.*x+y.*y+1);
%均值核
% kernel=ones(2*ds+1,2*ds+1);
kernel=kernel./((2*ds+1)*(2*ds+1));
dst=zeros(m,n);% output
%---------------------------%
%  Non-Local Means Denoising
%---------------------------%
for i=1:m
    for j=1:n
        
        %当前点坐标和邻域窗口
        i1=i+offset;
        j1=j+offset;
        W1=PaddedImg(i1-ds:i1+ds,j1-ds:j1+ds);
        
        %加权因子矩阵和图像
        weight=zeros(2*Ds+1,2*Ds+1);
        image=PaddedImg(i1-Ds:i1+Ds,j1-Ds:j1+Ds);

        for r=-Ds:Ds
            for s=-Ds:Ds
                
                %跳过当前点
                if(r==0&&s==0)
                    continue;
                end
                
                %待加权点坐标和邻域窗口
                i2=i1+r;
                j2=j1+s;
                W2=PaddedImg(i2-ds:i2+ds,j2-ds:j2+ds);
                
                %核加权的距离和加权因子
                distance=sum(sum(kernel.*(W1-W2).*(W1-W2)));
                weight(r+Ds+1,s+Ds+1)=exp(-distance/(h*h));
            end
        end
        
        %最大权重赋给当前点，归一化权重
        weight(Ds+1,Ds+1)=max(max(weight));
        weight=weight/(sum(weight(:)));
        
        dst(i,j)=sum(sum(image.*weight));
    end
end
%---------------------------%
%       output
%---------------------------%
subplot(121),imshow(g1,[]),title('待处理');
subplot(122),imshow(dst,[]),title('NL-means除噪');
psnr(zyy,uint8(g1))
psnr(zyy,uint8(dst))
ssim(zyy,uint8(g1))
ssim(zyy,uint8(dst))

4.均值、中值、高斯、维纳、小波、NL-means的PSNR、SSIM对比

	待除噪图	均值滤波	中值除噪	高斯除噪	维纳除噪	小波除噪	NL-除噪
PSNR	20.7648	23.724	24.2671	24.0838	26.0139	22.3159	26.8054
SSIM	0.3504	0.5863	0.5518	0.6006	0.7006	0.3979	0.6385