How to traverse along skeleton just once instead of multiple times?

Here is my code

    I=im2bw(I);
    skel = bwmorph(I,'thin',inf)
    boundaries = bwboundaries(skel);
    thisBoundary=boundaries{1};
    BL=size(thisBoundary(:,2));
    figure(1)
    imshow(skel)
    figure(2)
    imshow(I)
    hold on
    for j=1:1:BL(1)
        plot(thisBoundary(j,2), thisBoundary(j,1), '.c', 'LineWidth', 2);
        disp(["x: ",num2str(thisBoundary(j,2)), "y: ",num2str(thisBoundary(j,1))])
        pause
    end

and sample image

As it traverses along the skeleton, at the beginning, you will notice if you pay attention to the outputted x and y values, that as the green points go upwards on the image, eventually it traverses back down and hits the same x,y points again

For example, if you see the output

    x: 208 y:22
    x: 208 y:21
    x: 208 y:20
    x: 208 y:21 <-already hit this (x,y) pair 2 lines above
    x: 208 y:22 <-already hit this (x,y) pair 4 lines above

Is there a way to somehow sort the values in `thisBoundary` so that it only goes from one endpoint to the other endpoint once as opposed to hitting the same points multiple times?

Answer 1

[Image Analyst]

I just posted code for this a few days ago. Basically you can skeletonize, prune with the 'spur' option of bwmorph(), and then find branchpoints. and iterate until you have only two endpoints left. Search the forum for endpoint. If you can't find it or do it, let me know.

[Image Analyst]

Does it have to be exactly along the centerlines of the thick white bars? If not, use bwdistgeodesic(). See Steve's blog. http://blogs.mathworks.com/steve/2011/11/01/exploring-shortest-paths-part-1/

How did these things get so thick to begin with? Why are they not one pixel wide right from the start?

[Image Analyst]

Tell me what you want to do assuming you had the (x,y) coordinates of a path along the white bars.

[Image Analyst]

Then just call bwmorph(bw, 'skel', inf) and then call bwmorph() again with the spur option with a big enough factor to chop off any spurs that may be there.

[Image Analyst]

We've been working on an algorithm for extracting the longest spine in a crazy-looking skeleton with many arms, but the Mathworks is not quite ready to release it publicly yet. Sorry.