Why does imshow display different image when I try to show a CT image?

Hi,

I want to show an original abdominal CT (Computed Tomography) image (which is uint16 bit ) with these codes;

im = imread('Slice_87.png');
imshow(im);

The original image and the figure shown by the above command are different.

For instance, the figure shown by the above command has a grayscale background. However, the background in the original image seems as black when I open it with ImageJ or paint.

Why are they different?.

How can I show the CT image as it is seen by ImageJ or paint?

Answer 1

[KALYAN ACHARJYA]

Can you try this way-

[img,map] = imread('Slice_87.png');
imshow(img,map);

[S C.Carl]

Hi,

Thanks for your response. Your code imshow(img,map) shows the image as grayscale. Thanks.

I have one more problem:

I would like to get a grayscale image after the following codes;

onlyLiver_asGrayscale = uint8(BinaryMaskImg) .* uint8(wholeCTImg_asGrayscale);

figure(1); imshow(onlyLiver_asGrayscale);

However, the result image named as onlyLiver_asGrayscale is still binary.

Why is the image "onlyLiver_asGrayscale" still binary ?

[KALYAN ACHARJYA]

Can you please attach the sample image file?

[DGM]

If you're trying to apply the mask to the image, then the mask must be unit-scale. Unless the mask is binarized, it must also be floating-point. If it is required that the mask is floating point, then the image must also be floating point.

All throughout these conversions between classes, you either rescale your data according to the class, or you need to keep track of your data scale and make sure that it doesn't cause problems (e.g. truncation). If you're using uint8(), you aren't rescaling anything, and it's not possible for me to infer what the actual data scaling is, so chances are your data scaling is inappropriate and it's resulting in severe truncation.

inpict = imread('cameraman.tif'); % uint8-scale uint8
mask = imread('cmantifmk.png'); % uint8-scale uint8

% multiply two uint8-scale images
% nothing left but the mask due to severe truncation
outpict1 = mask.*inpict;
imshow(outpict1)

% multiply a unit-scale uint8 mask with a uint8-scale image
% works, but will break if mask or inpict aren't uint8
% also won't work if mask is not binarized
outpict1 = mask/255.*inpict;
imshow(outpict1) 

% multiply two unit-scale float images
% that works regardless of input class
% works for binarized masks and graduated masks
outpict3 = im2double(mask).*im2double(inpict);
outpict3 = im2uint8(outpict3); % cast _and rescale_
imshow(outpict3) 

[S C.Carl]

Thanks for your response.

However, your codes still show binary result rather than a grayscale result :(

Our images are attached.

How to solve this problem with them?

[KALYAN ACHARJYA]

Once I checked, it seems fine, where did you get the binary image from, as you mentioned.

More, you are requested to read @DGM comments again.

[Simon Chan]

Following DGM comments should work.

The issue may be due to the fact that the range of pixel value is too large when using function imshow.

By limiting the range to the pixel values within the mask, you should be able to get a btter image.

I=imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1382669/wholeCTgray_87.png');
mask = imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1382664/MASK_87.png');
outpict3 = im2double(mask).*im2double(I);
outpict3 = im2uint16(outpict3);                         % Your original image is uint16
[minValue,maxValue] = bounds(outpict3(outpict3~=0));
imshow(outpict3,[minValue,maxValue])                

[S C.Carl]

Thank you.

But, I want to save the grayscale result as a png or tiff file, and use;

imwrite(outpict3, name);

Then, the saved image still seems as black (not grayscale) when it is opened by Paint or ImageJ. I still could not save the result as grayscale :(

[Simon Chan]

One possible way is use function getframe as follows:

I=imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1382669/wholeCTgray_87.png');
mask = imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1382664/MASK_87.png');
outpict3 = im2double(mask).*im2double(I);
outpict3 = im2uint16(outpict3);                         % Your original image is uint16
[minValue,maxValue] = bounds(outpict3(outpict3~=0));
imshow(outpict3,[minValue,maxValue]);
F = getframe(gca);
imwrite(F.cdata,'ScreenCapture.png');

[DGM]

Use mat2gray() to do the normalization instead of relying on figure capture and all the problems that causes.

I=imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1382669/wholeCTgray_87.png');
mask = imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1382664/MASK_87.png');

outpict3 = im2double(mask).*im2double(I); % multiply
[minValue,maxValue] = bounds(outpict3(outpict3~=0)); % get data range
outpict3 = mat2gray(outpict3,[minValue,maxValue]); % normalize to extrema
outpict3 = im2uint16(outpict3); % Your original image is uint16

imshow(outpict3); % now it's scaled as expected

imwrite(outpict3,'ScreenCapture.png');