Why does plot(X,Y,'or') and plot(X,Y) give different results

So,I have a set of points that I am trying to visualize.

The points are obtained by running the code mentioned in the below link

https://in.mathworks.com/matlabcentral/fileexchange/48908-accurate-subpixel-edge-location

Basically the output of this is a structure that stores the positions of the edges along with some other stuff.

Now,when I plot edges.x and edges.y using the following command

plot(edges.x,edges.y,'or'); The result comes out like this

Whereas the result of plot(edges.x,edges.y) is like this

The original result of the code is as follows

Any idea why the line plot is coming out like this,and what can be done to correct it.

Note 1 : After inspecting the picture,I was able to visually figure out the correct order.If you look at edges.mat the expected starting point is [103,66.3015] and the same is the end point

Answer 1

[Walter Roberson]

This is to be expected. There is nothing that requires that the algorithm returns the point is edge-tracing order.

You should use boundary() https://www.mathworks.com/help/matlab/ref/boundary.html to detect the order

idx = boundary(edge.x, edge.y);
plot(edge.x(idx), edge.y(idx))

[Sparsh Garg]

hmm,well i tried the mentioned approach,but the result is not what I want.

I am assuming that it tried to fit the closest polygon to all points but in the end we want the points to resemeble a V,but the approach of using boundaries isn't really giving me what i want.

[Walter Roberson]

shrink = 0.75;
idx = boundary(edge.x, edge.y, shrink);
plot(edge.x(idx), edge.y(idx));

You might need to increase "shrink" (it has a maximum of 1.0)

[Sparsh Garg]

Well,this does work for the image mentioned in the post,but it doesn't work on other images

For example: original image

Point plot

Result of running plot(edges.x(idx),edges.y(idx))

[Walter Roberson]

You might need to increase shrink even more (maximum 1.0)

However, if you look near the 45 level, you can see there the closest approach between the two distinct blobs. Now compare that distance to the size of the gap in the middle of the base near 78 or so: the base gap looks close to the same or possibly even larger. So one time you want the distance to be inconsequential not indicating a different object, but the other time you want the distance to indicate a new object.

I thought for a second that perhaps you could get somewhere by taking the two nearest neighbours of each point, but look again at the gap at the base. The second tick mark from the left, the pair of points close together: the two nearest neighbours to the right-hand one of the pair are both to the left.

    ABCD EF    GHIJ

the two nearest neighbours of F are D and E, not E and G. So nearest neighbour techniques in themselves are not going to be enough. Perhaps you could get somewhere with roughly the 5 nearest neighbours and some logic.

Answer 2

[Chunru]

The line plot draws line from one point to another point. The order of points is thus important. The 'or' option just plots the data points without draw line.

If the data points were in right order, you could have seen a line plot more similar to the scatter plot ('ro').

Answer 3

[KSSV]

When you plot with markers (i.e. 'o'), plot will not connect the markers. But when you simply plot, it will try to join them by lines. Your points are not in order, so they are joined as the points are given.

Attach your points to get what you want.

[Sparsh Garg]

I guess that's what I will have to do eventually,but in the end,correcting the order of the poins manually won't fit my end objective of designing an algorithm that can generalize to any images.

For a complex image this is the result.And so reorderign the points (which I admit has to be done) will be painstakingly brutal.