image thumbnail

Divided edge bundling in Matlab

version (231 KB) by Kelly Kearney
Calculate and plot graph divided edge bundle pathways for directional network graphs


Updated 23 Sep 2021

From GitHub

View license on GitHub

Divided edge bundling for Matlab

Author: Kelly Kearney View Divided edge bundling in Matlab on File Exchange

This repository includes the code for the debundle.m and plotdeb.m Matlab functions, along with all dependent functions required to run them.

These two functions replicate the divided edge bundling algorithm, which provides an edge-routing method to declutter complex directional network graphs while highlighting major directional edge pathways. The algorithm, which is an extention of force-directed edge bundling, was developed by David Selassie and is described in the following paper:

  • Selassie D, Heller B, Heer J (2011) Divided edge bundling for directional network data. IEEE Trans Vis Comput Graph 17:2354-2363

Please note that I am not a coauthor on this paper; I simply coded this Matlab version so I could incorporate the divided edge bundling algorithm into my own research. While I would appreciate acknowledgement if you happen to use this particular implementation of the algorithm, please be sure to cite Salassie's paper as the primary citation.


  • Getting started
  • Syntax
  • Example
  • Contributions

Getting started


This function requires Matlab R2015b or later.

Downloading and installation

This code can be downloaded from Github.

Matlab Search Path

The following folders need to be added to your Matlab Search path (via addpath, pathtool, etc.):



Gnew = debundle(G) calculates bundled edge pathways based on the exisiting digraph G. The Node table in G must include Name, x, and y properties for each node, and the Edge table in G must include EndNodes and Weight for each edge. The output digraph Gnew will be identical to G but with x, y, and BundleCompat properties added to the Edge table.

Gnew = debundle(G, Name, Value, ...) adjusts the edge bundling calculation parameters. See function help for a list of named parameters.

plotdeb(Gnew) plots the edges of an edge-bundled digraph as color-transitioning patches with widths corresponding to bundle weight.

h = plotdeb(Gnew) returns the handle of the plotted patch object.


This example will use the 3 x 3 example from the Selassie 2011 paper (used in Figures 7, 8, and 9 of the paper). It consists of two columns of 6 nodes each, with edges crisscrossing in both directions.

We'll start with lists of node coordinates and names, as well as source and target node names for each edge:

x   = [  -50  -50  -50   50   50   50  -50  -50  -50   50   50   50];
y   = [    0  -10  -20    0  -10  -20  -30  -40  -50  -30  -40  -50];
id  = {  '0'  '1'  '2' '3b' '4b' '5b' '0b' '1b' '2b'  '3'  '4'  '5'};

src = {  '0'  '3'  '3'  '5'  '5'  '4' '3b' '4b' '5b' '4b' '1b' '0b' '4b' '2b'};
tar = {  '5'  '0'  '1'  '2'  '0'  '2' '1b' '1b' '1b' '0b' '4b' '5b' '2b' '5b'};

We can then use this information to build a Matlab digraph object.

G = digraph(src, tar, ones(size(tar)))

G = 

  digraph with properties:

    Edges: [14x2 table]
    Nodes: [12x1 table]

A regular Matlab digraph object doesn't include coordinates for its nodes, but the divided edge bundling algorithm will need this. We can add the coordinates as new table variables under the the Nodes property of the diagraph object:

[~,loc] = ismember(G.Nodes.Name, id);
G.Nodes = [G.Nodes table(x(loc)', y(loc)', 'variablenames', {'x','y'})];

disp('G Nodes table:');
disp('G Edges table:');
G Nodes table:

ans =

  12x3 table

    Name     x      y 
    ____    ___    ___

    '0'     -50      0
    '5'      50    -50
    '3'      50    -30
    '1'     -50    -10
    '2'     -50    -20
    '4'      50    -40
    '3b'     50      0
    '1b'    -50    -40
    '4b'     50    -10
    '5b'     50    -20
    '0b'    -50    -30
    '2b'    -50    -50

G Edges table:

ans =

  14x2 table

      EndNodes      Weight
    ____________    ______

    '0'     '5'       1   
    '5'     '0'       1   
    '5'     '2'       1   
    '3'     '0'       1   
    '3'     '1'       1   
    '4'     '2'       1   
    '3b'    '1b'      1   
    '1b'    '4b'      1   
    '4b'    '1b'      1   
    '4b'    '0b'      1   
    '4b'    '2b'      1   
    '5b'    '1b'      1   
    '0b'    '5b'      1   
    '2b'    '5b'      1   

Here's the resulting digraph, depicted using Matlab's built-in graph plotting function:

h = plot(G, 'XData', G.Nodes.x, 'YData', G.Nodes.y, ...
    'NodeColor', 'r', 'MarkerSize', 8);
axis equal;

Alternatively, we can plot the edges using plotdeb in initial mode, where the edges are just depicted as straight lines without any bundle weighting:

he1 = plotdeb(G, 'initial', true);
set(h, 'EdgeColor', 'none');
uistack(h, 'top');

cb = colorbar('south');
set(cb, 'position', [0.2 0.2 0.2 0.02]);
set(cb, 'ticks', [0 1], 'ticklabels', {'Source', 'Target'});

We can now run the divided edge bundling algorithm. The code runs pretty quickly for a small graph like this example, but it can take a while for the calculations to complete when used on large, complex graphs. The progress bars printed to the Command Window show the progress as the code preprocesses each edge and then runs 5 passes of the bundling calculation.

G = debundle(G);
 100% [=================================================>]   14/  14
 Elapsed time is 0.024743 seconds.
 Pass 1
 100% [=================================================>] Pass 1: 29 of 29
 Elapsed time is 0.047363 seconds.
 Pass 2
 100% [=================================================>] Pass 2: 29 of 29
 Elapsed time is 0.101606 seconds.
 Pass 3
 100% [=================================================>] Pass 3: 29 of 29
 Elapsed time is 0.213681 seconds.
 Pass 4
 100% [=================================================>] Pass 4: 29 of 29
 Elapsed time is 0.433725 seconds.
 Pass 5
 100% [=================================================>] Pass 5: 29 of 29
 Elapsed time is 0.884515 seconds.

We can plot the new bundled and weighted edges with the plotdeb function (and hide the old edges):

he2 = plotdeb(G, 'w', 5);
uistack(h, 'top');


Community contributions to this package are welcome!

To report bugs, please submit an issue on GitHub and include:

  • your operating system
  • your version of Matlab and all relevant toolboxes (type ver at the Matlab command line to get this info)
  • code/data to reproduce the error or buggy behavior, and the full text of any error messages received

Please also feel free to submit enhancement requests, or to send pull requests (via GitHub) for bug fixes or new features.

Published with MATLAB R2018a

Cite As

Kelly Kearney (2021). Divided edge bundling in Matlab (, GitHub. Retrieved .

MATLAB Release Compatibility
Created with R2017a
Compatible with any release
Platform Compatibility
Windows macOS Linux
Tags Add Tags

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!










To view or report issues in this GitHub add-on, visit the GitHub Repository.
To view or report issues in this GitHub add-on, visit the GitHub Repository.