Denoise Image with Gaussian Noise Using MATLAB / Octave

Question

I want to remove a noise for an image using MATLAB, when the observed image is $$f=u+v$$ where $u$ is the restored image (is the image i want recovered) and $v$ is the gaussian noise.

To restore $u$, I solve the following minimization problem: $$\min_{u \in H^1(\Omega)} \int_\Omega \gamma|\nabla u(x)|^2dx+ \int_\Omega (f(x)-u(x))^2dx,$$ where $\gamma$ is the regularization coefficient and $\Omega=[0,n]\times[0,m]$ and $[n,m]=$size($u$).

I want to solve the PDE (Euler-Lagrange) using MATLAB: \begin{eqnarray} div(\gamma \nabla u) + u = f \;in \;\Omega \\ \frac{\partial u}{\partial n}=0 \; in \; \partial \Omega \end{eqnarray}

Can anyone help me to solve this problem? Thank you!

I tried the following code :

clear all, close all,clc

uor=imread('gourd.bmp'); % the original image 
u0 = imnoise(uor,'gaussian',0,0.01);
u0=double(u0);
[m n]=size(u0);
uor=double(uor);
u=u0;
c=0.028;
h=1;     
for Iter=1:50, 
    for i=2:m-1,
      for j=2:n-1,
          Lap=0.003*(u(i+1,j)+u(i-1,j)-4*u(i,j)+u(i,j+1)+u(i,j-1));
          u(i,j)=(u0(i,j)+(1/(2*c*h*h))*Lap);
     end
    end
    for i=2:m-1,
          u(i,1)=u(i,2);
          u(i,n)=u(i,n-1);
        end

    for j=2:n-1,
          u(1,j)=u(2,j);
          u(m,j)=u(m-1,j);
        end

        u(1,1)=u(2,2);
        u(1,n)=u(2,n-1); 
        u(m,1)=u(m-1,2);
        u(m,n)=u(m-1,n-1);

en=0.0;  
    for i=2:m-1,
      for j=2:n-1,
      ux=(u(i+1,j)-u(i,j))/h;
      uy=(u(i,j+1)-u(i,j))/h;
      fidelity=(u0(i,j)-u(i,j))*(u0(i,j)-u(i,j));

      en=en+c*fidelity;
      end
    end


Energy(Iter)=en; 

%  Error between uor and u0
 ur=reshape(u,m*n,1);
 uori=reshape(uor,m*n,1);
 residu=norm(ur-uori)/norm(uori);

 [peaksnr, snr] = psnr(uor, u);

disp(['    iter ' num2str(Iter), ' :     Error = ' num2str(residu), ...
    ' ,    Peak-snr ' num2str(-peaksnr), ' ,    SNR ' num2str(snr)]);


  end 

% show the structural similarity index for measuring image quality 
[ssimval, mapssim] = ssim(u,uor);
disp([' the structural similarity index is ' num2str(ssimval)]);
figure,imshow(mapssim,[]); axis square; 

figure,imagesc(u); axis image; axis off; colormap(gray);

The original image is here : https://www.dropbox.com/s/4bccby1f4lxp4j9/gourd.rar?dl=0

Best regards

Answer 1

You're trying to solve what's called Perona Malik Non Linear Diffusion Problem (Sometimes people call it, by mistake, Anisotropic Diffusion).

Anyhow, the simplest code for that is Anisotropic Diffusion (Perona & Malik) on The MATLAB File Exchange.

There is a more advanced (Anisotropic for real) algorithm in Fast Anisotropic Curvature Preserving Smoothing (Also on the File Exchange - Fast Anisotropic Curvature Preserving Smoothing).

Enjoy...