Complex Networks, Simple Vision

Abstract

This paper proposes and illustrates a general framework to integrate the areas of vision research and complex networks. Each image pixel is associated to a network node and the Euclidean distance between the visual properties (e.g. gray-level intensity, color or texture) at each possible pair of pixels is assigned as the respective edge weight. In addition to investigating the therefore obtained weight and adjacency matrices in terms of node degree densities, it is shown that the combination of the concepts of network hub and \emph{2-}expansion of the adjacency matrix provides an effective means to separate the image elements, a challenging task in computer vision known as segmentation.