How to apply distance transform on labelled matrix

I have got a labelled matrix and want to apply distance tranform on the diffent different labels in the matrix. How can it be done in matlab?

Thanks

[Walter Roberson]

When you apply the distance transform, are the regions labeled with other labels to be considered unpassable or passable? Or, for example, are regions with a numerically greater label number to be unpassable while regions with numerically lower label number are to be considered passable ?

Do all of the results need to somehow be synthesized into one matrix, or can there be one output for each distinct label?

[Rizwan Khan]

Dear Roberson, 

I think you got my question correctly, I need the results into one matrix.
I am attaching my code and flowchart here for clearity.

CC = bwconncomp(B,8);

L = labelmatrix(CC);

num = max(L(:));

D = zeros(size(B));

for iter = 1:Inf % loop for new regions

for j= 1:num

[r,c] = find(L==j);

for k = 1:size(r,1)

D = my_dist(B(r(k,1),c(k,1)));

end

L = watershed(D);

num1(j) = max(L(:));

end

num = sum(num1);

if abs(num(iter+1)-num(iter)) == 0

break

end

end

[Walter Roberson]

is this https://ieeexplore.ieee.org/document/885672

[Rizwan Khan]

I am referring this paper: https://www.icevirtuallibrary.com/doi/10.1680/jgeot.16.P.278

Answer 1

[Anmol Dhiman]

Hi Rizwan,

I am assuming the matrix as binary. You can refer to Distance Transform of a Binary Image.

Regrds,

Anmol Dhiman

[Walter Roberson]

The user specifically talks about a label matrix. Those are not binary (logical): they are numeric matrices in which each unique value defines a different region (except that 0 is always background.)

[Image Analyst]

That's why it doesn't make sense. Why on earth would you even want to try to do a distance transform on a labeled image? What would even be the formula for distance in that case? And what would he do with the result? I think this is a case of a user thinking he wants something, but he really doesn't want that, he just doesn't know it yet.

[Rizwan Khan]

Dear Image Analyst,

This is the flowchart which i need to implement in matalb.

CC = bwconncomp(B,8);

L = labelmatrix(CC);

num = max(L(:));

D = zeros(size(B));

for iter = 1:Inf % loop for new regions

for j= 1:num

[r,c] = find(L==j);

for k = 1:size(r,1)

D = my_dist(B(r(k,1),c(k,1)));

end

L = watershed(D);

num1(j) = max(L(:));

end

num = sum(num1);

if abs(num(iter+1)-num(iter)) == 0

break

end

end