Straightening image from X-ray tomography

Dear all,

I am scanning some no rigid materials with X-ray tomography. I got a stack of images where unfortunately is impossible to have the sample straight like you can see below.

I already tried to do it with ImageJ for one image and the result is not that bad.

The problem is: I have a stack of images with 1500 images and it's really long to do it with ImageJ. Do you have an idea of to straight the material for a stack of images with Matlab?

Best

Answer 1

[jonas]

Here is something I have stitched together that seems to work fairly well.

%%Load Image, rotate and convert to binary
  GRAY=imread('scan_00123.tif');
  GRAY=rot90(GRAY);
  threshold = graythresh(GRAY);
  BW = im2bw(GRAY, threshold);
%%Find Coordinates of images and smooth a curve
  [y,x]=find(BW==1);
  y=y-mean(y);
  figure;
  subplot(1,2,1);hold on
  plot(x,y)
  win=1500;
  y2=smooth(x,y,win);
  plot(x,y2,'r')
%%Fix size of vector
  [x, index] = unique(x); 
  xq=1:size(GRAY,2);
  yq = interp1(x, y2(index), xq,'linear','extrap'); 
  plot(xq,yq,'--k')
%%Adjust image column-wise
  NewImage=uint16(nan(size(GRAY,1)*3,size(GRAY,2)));
  ymid=size(GRAY,1)*3/2;
  for i=1:size(GRAY,2)
      s=yq(i);
      NewImage(ymid-s:ymid-1-s+size(GRAY,1),i)=GRAY(:,i);
  end 
  figure;
  NewImage(isnan(NewImage))=0;
  imshow(NewImage)

[redmef]

Thanks Jonas! It works for one image, but not for the first image of the stack for instance. I got "Subscript indices must either be real or positive intergers or logicals' Error: NewImage(50-s:49-s+size(GRAY,1),i)=GRAY(:,i);

Best

[jonas]

Yes, if you give my a few images I can generalize the code. The code I posted was tailored for the size of the image you gave. I think 1-2 more images would be sufficient.

[redmef]

Dear Jonas,

Please find attached a stack of the images.

Best Regards

[redmef]

An other question: it seems that the output image is converted into 8 bits. Is it possible to get it in 16bits.

[jonas]

I have updated the code. The issue was that the actual images are vertical, while the one I tested on initially was horizontal. Easily fixed by rot90. The image is now saved as uint16.

Btw, I successfully ran the code on all 4 images. However, to my untrained eye they are identical?

[jonas]

So did it work out?

[redmef]

Hello Jonas,

Sorry for the delau. It works perfectely. Thank you very much

[jonas]

Happy to help! Accept the answer please!

[redmef]

Just one quick question: this code is specific for this material and that direction; Does it work for any direction and thin material ?

[jonas]

'specific for this material?'

In general, no. Most of the script should work for any thin material. You may have to adjust the segmentation slightly depending on the color/brightness of the material.

specific direction?

The script, as is, works for materials aligned vertically. Simply remove the line

GRAY=rot90(GRAY);

for materials stretched horizontally.