Image segmentation by means of complex networks centrality indices
Caterina Fenu · 2021
In this paper, we propose a novel approach to perform multi-label segmentation. Starting from an image, we construct the associated weighted graph and assign to a small number of pixels (seeds) a label. The aim is to assign each unlabeled pixel to a label in order to identify different regions. Our algorithm uses the notion of communicability from complex networks theory to compute the easiness for an unlabeled pixel to reach a labeled one. By assigning each pixel to the label for which the greatest communicability is calculated, we can perform good image segmentation.