Weighted adjacency matrix of an image using energy function

Creates a weighted adjacency matrix from an energy matrix for an image. Average, similarity, and dissimilarity between the nodes i,j.

Weighted adjacency matrix of an image

Creates a weighted adjacency matrix from a energy matrix (E) for an image. Average, similarity and dissimilarity between the nodes i,j can be computed. If a custom energy matrix is provided, the same functions can be used to fill the adjacency matrix with edge weights.

Graphs table

Type	Images
4-noded unidirection
4-noded bidirection
8-noded, six edges, vertical and cross directions
8-noded, eight edges, unidirection
8-noded, eight edges, bidirection

Requirements

MATLAB

Run command

Please use the demo.m to run the program.

Change the hyper parameters accordingly if needed.

% Inputs
h = 3; w = 6;                       % Image size [height x width]
edgeDirection = 1;                  % Edge direction 1 - uni | 2 - bi
noded = 8;                          % Nodes 4-noded, 6-direction, 8-noded (pixels)
weight_type = 'Average';            % Average (E(i), E(j)) / 2 
                                    % Similarity (E(i) - E(j)) 
                                    % Dissimilarity 1 / (E(i) - E(j)).
                                    % Where E is the energy at node i, j.
flowDirection = 'col_wise';         % 'row_wise' | 'col_wise'

Known issues

1. Sometimes plot(G) or plot(G,'Layout','force') produces strange looking graphs.

Adaptation of open source

Some of the MATLAB adjacency functions are adapted from the Stack Overflow.

Feedback

Please rate and provide feedback for the further improvements.