How to resize an MRI image data keeping the original field of view?

I have few MRI datasets, where the dimension of the dataset is 320x640x16. First two dimensions are row and column and the last one is Coil dimension. I want to resize the row and column into 256x256x16, keeping the original Field of view. I tried the below code. But it make changes into the field of view and it cut the slices.

new_data = zeros([256,256,16]);
load brain.mat;
new_data(:,:,:) = raw_data(33:288, 193:448, :);

Answer 1

[Image Analyst]

First get one of the images then call imresize.

[rows, columns, numberOfSlices] = size(raw_data)
new_data = zeros(256, 256, numberOfSlices, class(raw_data));
for k = 1 : numberOfSlices
    thisImage = raw_data(:, :, k);
    new_data(:, :, k) = imresize(thisImage, [256, 256]);
end

[Gulfam Saju]

I used this code, but the image lost it's original data and became distorted.

[Gulfam Saju]

original image

[Image Analyst]

Looks like some kind of demosaicing problem or maybe your image is not really grayscale. Basically do whatever you have to do to get your good-looking original image and then call imresize(). Calling imresize on that image will not cause it to look like the spotted image you showed. If you still have trouble, attach a .mat file. If it's bigger than 5 MB, then write out one that has only a few slices contained in it.

[Gulfam Saju]

Use absolute value to see the image. This one has different dimension 320*320.

[Image Analyst]

Not sure where you got that weird looking image but this works fine

s = load('brain.mat');
raw_data = s.raw;
[rows, columns, numberOfSlices] = size(raw_data)
new_data = zeros(256, 256, numberOfSlices, class(raw_data));
for k = 1 : numberOfSlices
	thisImage = raw_data(:, :, k);
	subplot(2, 1, 1);
	imshow(thisImage, [])
	axis('on', 'image');

	new_data(:, :, k) = imresize(thisImage, [256, 256]);
	subplot(2, 1, 2);
	imshow(new_data(:, :, k), [])
	axis('on', 'image');
end

The original and resized images are shown below:

However you only put one 2-D image into the .mat file, not a 3-D image.

Answer 2

[Matthew Pepich]

I was too slow with my answer and @Image Analyst did it better, but here is another option that just samples the image. This works for displaying a thumbnail where interpolation is not important, but "imresize" is preferred if you need accuracy.

% Get your original image
C = imread('landOcean.jpg');

% Produce a scaled down image
scale = 0.10;
x = round( linspace( 1, size(C,1), size(C,1)*scale ) );
y = round( linspace( 1, size(C,2), size(C,2)*scale ) );
C2 = C(x, y, :);

% Display results
figure();
subplot(2,1,1);
image(C);
axis image;
title( sprintf('Size = %d x %d',size(C,[1 2])) )
subplot(2,1,2);
image(C2);
axis image;
title( sprintf('Size = %d x %d (Scaled to %g%%)',size(C2,[1 2]), 100*scale) )

[Gulfam Saju]

actually, its not an image file. These are complex data types converted into "mat" files from h5 files.

[Matthew Pepich]

Would it still work to use "linspace" to generate even indices? Ignoring the display part of my solution, try just using this to produce your indices:

x = round( linspace( 1, size(raw_data,1), 256 ) );
y = round( linspace( 1, size(raw_data,2), 256 ) );
new_data = raw_data(x, y, :);