why I fully lose dicom image after adding noise to it?

Dear all,

I am trying to add both Gaussian and salt & pepper noise to a dicom image using imnoise() but in any densities I completely lose the image it becomes fully black with white spots in it.

I would be glad if someone could help me.

The below is my image and my code.

img = dicomread('1.dcm');

density = 0.01;

%img = imnoise(img,'gaussian',density);

img = imnoise(img,'salt & pepper',density);

Answer 1

[Simon Chan]

imnoise expects pixel values of data type double and single to be in the range [0, 1]. You can use the rescale function to adjust pixel values to the expected range. If your image is type double or single with values outside the range [0,1], then imnoise clips input pixel values to the range [0, 1] before adding noise.

J = rescale(img);
img_noise = imnoise(J,'salt & pepper',0.01);

[khwaja mustafa siddiqi]

Thank you

[Simon Chan]

No worry. What you need is an additional work on scaling.

Following code shows the scaling using the maximum and minimum values from the original DICOM image.

The subtraction of two images (final vs original) in class uint16 shows the noise only, so the original image information is still there,

I think there are some other methods to achieve the same result, but may be I am too lazy and still use function rescale to do this. :-)

clear; clc;
filename = unzip('https://www.mathworks.com/matlabcentral/answers/uploaded_files/910755/1.zip');
I = dicomread(filename{1});
maxValue = double(max(max(I)));
minValue = double(min(min(I)));
%
J = rescale(I);
Jnoise = imnoise(J,'salt & pepper',0.01);
Irestore = uint16((Jnoise.*(maxValue-minValue))+minValue);
subplot(2,2,1)
imshow(I,[]);
title('Original DICOM Image');
subplot(2,2,2)
imshow(Jnoise,[]);
title('Rescale Image with noise');
subplot(2,2,3)
imshow(Irestore,[])
title('Image back to class uint16');
subplot(2,2,4)
Idiff = imabsdiff(Irestore,I);
imshow(Idiff,[]);
title('Subtraction of Image');

[khwaja mustafa siddiqi]

Thank you, it helped me to achive what I wanted

[ABIRAMI RAMALINGAM]

thank you, i m also have this doubt.

Answer 2

[Walter Roberson]

class() has probably changed.

im2double(img) and imnoise() the results and im2uint8 or as appropriate to return to the original type.

There is a possibility that your image is int16 with signed data, especially if it is CT, and it might take a slight bit more work to get back to signed

[khwaja mustafa siddiqi]

my image is uint16, and it's MRI image

[Walter Roberson]

salt and pepper noise sets pixels to either their minimum or their maximum. Suppose that your useful data is in the range 4000 to 12000, but you add noise that is either 0 or 65535, then if you use imagesc() or imshow() with the [] option to rescale your image, then instead of the plot being based omn 12000-4000 = 8000 values, it is suddenly based on 65536 values, and the useful data would get washed out.

img = sort(sort(randi([1000 2000], 48, 64, 'uint16'),1),2);

density = 0.01;
%img = imnoise(img,'gaussian',density);
img2 = imnoise(img,'salt & pepper',density);

whos

  Name          Size            Bytes  Class     Attributes

  density       1x1                 8  double              
  img          48x64             6144  uint16              
  img2         48x64             6144  uint16              

[min(img(:)), min(img2(:))]

ans = 1×2

1000 0

[max(img(:)), max(img2(:))]

ans = 1×2

2000 65535

imshow(img)

imshow(img2)

[khwaja mustafa siddiqi]

what solution do you suggest me in order to work both with the gaussian and salt & pepper noises ?

[Image Analyst]

Since your images are in the range 0-256, I'd just immediately cast them to uint8 right after you read them in

img = dicomread('1.dcm');
img = uint8(img);

After that, everything should be fine.

[khwaja mustafa siddiqi]

Thank you, It worked just fine

Answer 3

[Image Analyst]

Try this:

img = dicomread('1.dcm');
subplot(2, 2, 1);
imshow(img, []);
subplot(2, 2, 2);
imhist(img);
% Show min and max are 0 and 256, 
% nowhere close to the uint16 range of 0 to 65,535.
min(img(:))
max(img(:))
density = 0.01;
%img = imnoise(img,'gaussian',density);
noisyImage = imnoise(img,'salt & pepper',density);
whos noisyImage
subplot(2, 2, 3);
imshow(noisyImage, [])
% Now rescale
noisyImage = imnoise(mat2gray(img),'salt & pepper',density);
subplot(2, 2, 4);
imshow(noisyImage, [])

[khwaja mustafa siddiqi]

Thank you