% test_wa2_denoising.m -% image denoising using wave atoms and waveletsdwtmode('per');for runid=1:2  if(runid==1)    x = double(imread('data_seismic.jpg'));    sigma = 0.1;  else    x = double(imread('data_fingerprint.jpg'));    sigma = 0.2;  end    if(length(size(x))==3)    x = x/max(x(:));      x = rgb2gray(x);  end  x = (x-min(min(x)))/max(max(x-min(min(x))));    N = size(x,1);  noisy_x = x + sigma * randn(size(x));    figure; clf; imagesc(x); axis equal; axis tight; colormap gray; colorbar;  title('input image');    figure; clf; imagesc(noisy_x); axis equal; axis tight; colormap gray; colorbar;  title('noisy image');    threshold = 3 * sigma;  shifts = [1 2 4];    %wave atom frame denoising  pat = 'p';  tp = 'directional';  val = zeros(size(noisy_x));  for is=shifts    for js=shifts      xx = circshift(noisy_x,[is,js]);      cfs = real(fwa2sym(xx,pat,tp));      cfs = cfs .* (abs(cfs)> sqrt(2)/2*threshold);      tmp = real(iwa2sym(cfs,pat,tp));      tmp = circshift(tmp,[-is,-js]);      val = val + tmp;    end  end  val = val / length(shifts)^2;  figure; clf; imagesc(val); axis equal; axis tight; axis off; colormap gray;  title(sprintf('wave atom frame, psnr = %.2f', -10*log10(sum(sum((val-x).^2))/(N*N))));    %wavelet denoising  val = zeros(size(noisy_x));  for is=shifts    for js=shifts      xx = circshift(noisy_x,[is,js]);      [cfs,s] = wavedec2(xx, log2(N), 'db5');      cfs = cfs .* (abs(cfs)>threshold);      tmp = waverec2(cfs,s, 'db5');      tmp = circshift(tmp,[-is,-js]);      val = val + tmp;    end  end  val = val / length(shifts)^2;  figure; clf; imagesc(val); axis equal; axis tight; axis off; colormap gray;  title(sprintf('wavelet db5, psnr = %.2f', -10*log10(sum(sum((val-x).^2))/(N*N))));end