Markov Random Fields Up:Graph
of line networks
An abstract representation of the studied features allows
us to only keep the relevant information for the tackled problem.
is defined by:
For our problem:
a set X of nodes.
a set U of edges.
Eventually a third set A which represents
the attributes corresponding to each node.
on cartographic data:
X : set of road segments
on image data:
U : edge between connected segments
A : length, orientation, localization (for the first
X : 1st step: road pixels, 3rd step: chains
of road pixels
U : 1st step: neighbor pixels (8 connectivity), 3rd step:
A : 1st step: localization, 3rd step: length, orientation
Christine Hivernat & Xavier Descombes