Given an image, it is possible to represent it as a complex network. This can be done by establishing links between pixels/regions which have similar local features (e.g. gray-level, local texture, color, etc). Such representations are richer than traditional images because they incorporate explicit information about relationships between different image parts. In addition, they allow vision problems to benefit from the rich developments in complex networks research.
We have been revisiting several image analysis methods by considering the enhanced representations of images as networks. For instance, the segmentation of objects in images becomes a problem of community finding in complex networks, and textures can be characterized in terms of measurements of the topology of the respective networks.
Conversely, image analysis methods can conversely be applied to the characterization of geographical complex networks (i.e. networks whose nodes have defined spatial positions).