其他

开发平台：

Matlab

dt_LAWMLShrink.m：源码内容

							function y = dt_LAWMLShrink(x)
% Local Adaptive Image Denoising Algorithm
% Usage :%        y = dt_LAWMLShrink(x)
% INPUT :%        x - a noisy image
% OUTPUT :%        y - the corresponding denoised image
% Set the windowsize and the corresponding filter
windowsize  = 7;
windowfilt = ones(1,windowsize)/windowsize;
% Number of Stages
J = 6;
I=sqrt(-1);
% symmetric extension
L = length(x); % length of the original image.
N = L+2^J;     % length after extension.
x = symextend(x,2^(J-1));
% Forward dual-tree DWT
% Either FSfarras or AntonB function can be used to compute the stage 1 filters  
[Faf, Fsf] = FSfarras;
%[Faf, Fsf] = AntonB;
[af, sf] = dualfilt1;
W = cplxdual2D(x, J, Faf, af);
%W = normcoef(W,J,nor);%-------------------------
% Noise variance estimation using robust median estimator..
tmp = W{1}{1}{1}{1};
Nsig = median(abs(tmp(:)))/0.6745;
for scale = 1:J-1
    for dir = 1:2
        for dir1 = 1:3            
            % Noisy complex coefficients
            %Real part
            Y_coef_real = W{scale}{1}{dir}{dir1};
            % imaginary part
            Y_coef_imag = W{scale}{2}{dir}{dir1};
                                   
            % Signal variance estimation
            Wsig = conv2(windowfilt,windowfilt,(Y_coef_real).^2,'same');%---用Gauss分布的ML估计系数方差----
            Ssig = sqrt(max(Wsig-Nsig.^2,eps));          
            Y_coef = Y_coef_real+I*Y_coef_imag;
            
            Y_coef = LAWMLShrink(Y_coef,Ssig,Nsig);
            
            W{scale}{1}{dir}{dir1} = real(Y_coef);
            W{scale}{2}{dir}{dir1} = imag(Y_coef);
            
        end
    end
end
% Inverse Transform
%W = unnormcoef(W,J,nor);------------------
y = icplxdual2D(W, J, Fsf, sf);
% Extract the image
ind = 2^(J-1)+1:2^(J-1)+L;
y = y(ind,ind);

						