Search code examples
imagematlabimage-processingcomputer-visionconnected-components

Group numbers in a certain way in MATLAB

I finding the connected components in an image without using the computer vision toolbox in Matlab.

In the first scan of the given image, I have given labels to each pixel. I also have created a table that gives me the list of the components that are connected to each other. The table looks like this:

1   5
1   5
2   4
3   5
8   5
2   6

This table implies that all the pixels with labels 1 and 5 are connected,2 and 4 are connected, 3 and 5 are connected and so on. Also, since 1 and 5 are connected and 3 and 5 are also connected, this implies 1 and 3 are also connected.

I want to find a way to group all these components together. Is there any possible way?? Or some way that every connected pixel gets a label equivalent to the label with the minimum value i.e. if 4, 7 and 12 pixels are connected, all the pixels with labels 4, 7 and 12 should get a label of value 4.

Solution

  • It appears that you want to compute the connected components of your graph. This can be done fairly easily in MATLAB without any additional toolboxes.

    e = [1   5; ...
     1   5; ...
     2   4; ...
     3   5; ...
     8   5; ...
     2   6];
% MATLAB doesn't support repeated edges so remove them
e = unique(sort(e,2),'rows');
% create a graph
G = graph(e(:,1),e(:,2));
% plot the graph (optional)
plot(G);
% find connected components
bins = conncomp(G);
% get the indicies of nodes in each cluster
cluster = arrayfun(@(v)find(bins==v),1:max(bins),'UniformOutput',false);

    Results

    enter image description here

    >> cluster{1}
ans = 1     3     5     8
>> cluster{2}
ans = 2     4     6
>> cluster{3}
ans = 7