deblurring with wiener filter

i m trying to blur an image and then deblur it using a wiener filter that i created. but the blurred image shows unexpected results. it appears the image is getting divided into blocks which are then arranged out o order. what's the issue.?

i am trying to create a wiener filer to blur the image. i use a low pass LSI filter and then add gaussian noise to the blurred image. i assume the expected noise to be white and hence has a flat spectral density.

but before i can create by wiener filter to deblur it, the blurring filter(h) gives unexpected results. i am highlighting in bold letters the portion i assume is creating the problem. will the code not workfor those images for which number of rows is not eual to number of columns? in any case, its not working i guess when rows = columns either......please help me in correcting the error. in my code:

h=low pass filter
Snn=noise power
Suu = signal power

Code:

% function wien2(name,xdim,No,Snn)

%
% name = 'input image' (as in 'lenna.256')

% xdim = x dimension of input image (usually 256 or 512)

% No   = Variance of Noise

% Snn  = Power Spectral Density of Noise (Assumed White)

%
% This function takes an input image, runs it through a LSI filter h,
% and adds Gaussian noise to it.  The MSE between the degraded and
% original image is then calculated.  Weiner Filtering is then performed
% (using known h) and the degraded image is restored using
% the filter.  The MSE between the restored and original image is then
% calculated and returned.  Snn is set to a constant (assumes white noise)
%

No = 80;

% Load and Plot image

pict = imread('C:\Documents and Settings\All Users\Documents\My Pictures\Matlab pics\planet.png');

pict=rgb2gray(pict)
;
[xdim xdim]=size(pict)
;

clf

subplot(221)

imagesc(pict);

colormap(gray);

txt = ' before degradation';

title(txt)

axis square

axis off

***% Create LSI degradation model, need it to be phaseless

hi = 3.5^(-2);

h = zeros(256);

xl = 4;

xh = xdim - xl + 2;

h(1:xl,1:xl) = hi;

h(xh:xdim,1:xl) = hi;

h(1:xl,xh:xdim) = hi;

h(xh:xdim,xh:xdim) = hi;

% Create Gaussian noise, mean = 0, variance comes from input (No)

noise = sqrt(No)*randn(xdim,xdim);

% Run image through LSI Filter and then add noise

dpict = imfilter(pict,h);

dpict =double(dpict)+noise;

% Plot degraded image

subplot(222)

imagesc(dpict);

colormap(gray);

txt = [' with additive Gaussian Noise (mean=0, var=' , num2str(No), ')'];

title(txt)

axis square

axis off

Answer 1

[Image Analyst]

I don't understand what h is. It's just a big image of zeros with values padded around the edge. What's the point of that? Correcting the line

[xdim xdim]=size(pict) ;

to

[rows columns] = size(pict);

and then fixing up h to use rows and columns won't help the fact that h is weird. Your h is not a low pass filter like you wanted. For one thing it's the same size as the image and should be a lot smaller unless you wanted to totally, and I mean totally, erase all image structure, which is why your left side of the image is nothing but uniform noise.

Try it like this:

clc;
close all;
fontSize = 20;

% Read in a standard MATLAB color demo image.
folder = fullfile(matlabroot, '\toolbox\images\imdemos');
baseFileName = 'peppers.png';
fullFileName = fullfile(folder, baseFileName);
% Get the full filename, with path prepended.
fullFileName = fullfile(folder, baseFileName);
if ~exist(fullFileName, 'file')
  % Didn't find it there.  Check the search path for it.
  fullFileName = baseFileName; % No path this time.
  if ~exist(fullFileName, 'file')
    % Still didn't find it.  Alert user.
    errorMessage = sprintf('Error: %s does not exist.', fullFileName);
    uiwait(warndlg(errorMessage));
    return;
  end
end
rgbImage = imread(fullFileName);
% Get the dimensions of the image.  numberOfColorBands should be = 3.
[rows columns numberOfColorBands] = size(rgbImage);
% Display the original color image.
subplot(2, 2, 1);
imshow(rgbImage, []);
title('Original Color Image', 'FontSize', fontSize);
% Enlarge figure to full screen.
set(gcf, 'units','normalized','outerposition',[0 0 1 1]);
pict=rgb2gray(rgbImage) ; 
[rows columns numberOfColorChannels] = size(pict) ;
subplot(221)
imagesc(pict);
colormap(gray);
caption = 'Before degradation';
title(caption, 'FontSize', fontSize);
axis off

% Create Gaussian noise, mean = 0, variance comes from input (No)
No = 80;
noise = sqrt(No)*randn(rows,columns);
subplot(2, 2, 2);
imshow(noise, []);
title('Pure Noise', 'FontSize', fontSize);

% Run image through low pass filter.
windowSize = 15;
h = ones(windowSize, windowSize)/windowSize^2;
% Blur the image  by applying the low pass filter.
blurredImage = imfilter(pict,h);
subplot(2, 2, 3);
imshow(blurredImage, []);
title('Low pass filtered image', 'FontSize', fontSize);

% Add Gaussian noise.
degradedImage =double(blurredImage)+noise;
% Display degraded image
subplot(2, 2, 4)
imagesc(degradedImage);
colormap(gray);
caption = sprintf('Degraded image with additive\nGaussian Noise (mean=0, var=%f)', No);
title(caption, 'FontSize', fontSize);
% axis square
axis off

[shawn]

but the context i read this method from require the use of a zero phase filter as a blurring filter.... how to achieve that???

[Image Analyst]

I don't know what that is by that specific name. A simple blurring filter centered in the window (kernel) will not cause any shift in the image.