MATLAB: Gaussian Noise and mean filter:

denoising imagesimage processingImage Processing ToolboxMATLABnoiseremoving noise

Hello Dear Experts,

If I am given a picture with pre-added Gaussian noise, and I know the mean and the var parameters. I need to use a best mask to enhance the image by removing the noise.

I know it should be a matrix 3×3 or 5×5 divided by the sum of the elements.

How should I solve such a problem?

Best Answer

That would not be the best method. In fact, it's often one of the worst. Nonetheless I gave some homework hints to Thomas who asked the same thing yesterday in http://www.mathworks.com/matlabcentral/answers/36869-i-need-help-with-matlab-assignment

noiseReducedImage = conv2(greenChannel, kernel);

kernel could be an N by N box with all ones (hint: use the ones() function). Make sure the kernel elements sum to 1 or else you're going to change the mean brightness of the image (so divide by N^2). Convolution is linear filtering. You could also use imfilter to do almost the same thing. imfilter is supposedly a little faster (according to the developer) and it doesn't flip the kernel like convolution does, though that doesn't make any difference if your kernel is symmetric.

If this is not homework and you need more guidance, let me know. The whole thing is just one line of code though and I said how to do it above.

Related Solutions

MATLAB: Quick simple question, mean filter

Don't use bad names like I and J. Pick descriptive names that aren't easily confused with 1 (one) or the imaginary variable. So, taking that advice:

windowWidth = 5;
kernel = ones(windowWidth) / windowWidth ^2;
outputImage = conv2(inputImage, kernel);

MATLAB: I need help with Matlab assignment

You didn't extract the green channel, you set the red channel and blue channel to zero. That's not the same. Here, use this code:

% Extract the individual red, green, and blue color channels.
redChannel = rgbImage(:, :, 1);
greenChannel = rgbImage(:, :, 2);
blueChannel = rgbImage(:, :, 3);

To do the filter, you'd use intlut() if you want to change the image values.

map = gray(256);
map(1:slow) = slow;
map(shigh:    You finish the rest!

To change the gray levels:

newGreenChannel = intlut(You finish the rest!

or to just change the display of it without changing the actual image itself,

imshow(greenChannel);
colormap(You finish the rest!
colorbar;

I've left parts undone for you to finish because that's what you requested. For noise reduction, you can blur the image with conv2().

noiseReducedImage = conv2(greenChannel, kernel);

kernel could be an N by N box with all ones. Make sure the kernel elements sum to 1 or else you're going to change the mean brightness of the image (so divide by N^2). Convolution is linear filtering. You could also use imfilter to do almost the same thing. imfilter is supposedly a little faster (according to the developer) and it doesn't flip the kernel like convolution does, though that doesn't make any difference if your kernel is symmetric.

Best Answer

Related Solutions

MATLAB: Quick simple question, mean filter

MATLAB: I need help with Matlab assignment

Related Question