Abstract
The theory and tools of Complex Networks have not been much applied to Image Analysis and Computer Vision problems. This paper introduces a new method for detecting interest points in digital images making use of Complex Network Analysis. This analysis includes a self-consistent post-processing procedure that improves the localization of the initially detected interest points in the image. We propose a general post-processing localization method based on centrality measures on a weighted version of the line-graph L(G) after the association of a spatial and weighted complex network G to each image with a prescribed geometrical structure. The practical testing of this fast-computable post-processing method shows that it is self-consistent since the distribution of the centrality measures in the weighted line-graph give us the intrinsic thresholds of interest of each region in the digital image.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
