How to detect a curve (surface roughness) from image?

I am working with a stack of 2D projections from X-ray microCT. I am going to find the coordinates of surface roughnesses curvature on 2D images and then extend it to 3D to have a cloud of points.

Images are like this:

I used "edge" function with different methods, but the results are not perfect, there are some separated lines and noises. I appreicate any better solution for this purpose and how can I get the coordinates of the detected curve? I assume then I have to apply a loop to get the results for the whole volume.

Thank you in advance for your help/comments.

Answer 1

[darova]

• binarize image first
• remove noise using bwareaopen
• use edge
• remove noise if needed using bwareaopen

[Sepehr Simaafrookhteh]

Thank you darova. That worked.

Now I applied the same procedure in a loop to a stack of 2D images. To find the coordinates of detected edge (I), I used "find" function:

[y, x]=find(I);

But the size of y and x matrix is different for each 2D slice. Hence, I cannot save all coordinates together in a 3D matrix. Could you please help me here?

[darova]

Try interp1 to make all x and y the same size

data = zeros(n,100);       % n is a number of picture
for i = 1:n
    data(i,:) = interp1(x,y,1:length(x));
end
surf(data)

[Sepehr Simaafrookhteh]

Thank you very much darova. But I received this error:

Error using matlab.internal.math.interp1
Sample points must be unique and sorted in ascending order.
Error in interp1 (line 154)
VqLite = matlab.internal.math.interp1(X,V,method,method,Xqcol);

But imresize worked to resize them to a single size for all projections. Is this a correct approach? However yet I am not sure how to save thousand of [x y] for each image (z) to have a cloud of points. I used xlswrite to save to an Excel file, but this ways I could only save [x y] for each image on separated excel sheets.

[darova]

• But imresize worked to resize them to a single size for all projections. Is this a correct approach?

Are the images have the same size?

• However yet I am not sure how to save thousand of [x y] for each image (z) to have a cloud of points

Why do you need x variable? It's always the same: [1 2 3 ...]. I'd just save y variable as 2D matrix

[Sepehr Simaafrookhteh]

Thank you for your responce. Yes, the input images are in the same size.

This is the output detected edge:

Not coonected everywhere, but the best I could receive so far. And here is the first values of the corresponding [x y] using find function:

It is a 2564x2 double matrix:

I need to save these changing coordinates for each image. Let's say 500 slices.

[darova]

Let's say A is your image

A = sprintf('image.png');        % read first image to get size
[m,n] = size(A);
ix1 = 1:n;                       % value for interpolation
Z = zeros(500,n);                % create matrix for 500 images profiles
for i = 1:500
    s = sprintf('image%d.png',i);
    A = imread(s);      % read binary image
    [~,y] = max(A);     % find max index (ones)
    ix = find(y);       % find non-zero elements
    y1 = interp1(ix,y(ix),x1);   % inteprolate zero values
    Z(i,:) = y1;        % fill 2d matrix
end
surf(Z)

[Sepehr Simaafrookhteh]

Thank you darova. You really helped me with your hints.