Common Visual Pattern Recognition Using Hierarchical Clustering Methods with Asymmetric Networks

MAGDALENA BOUZA; BRUNO CERNUSCHI FRÍAS

Rosario, Santa Fe

Congreso; Actas de ASAI 2015, 16º Simposio Argentino de Inteligencia Artificial, 31 de Agosto al 4 de Septiembre 2015.; 2015

SADIO, Sociedad Argentina de Informática

In this paper we propose a new method for common visualpattern identification via Directed Graphs. For this we match commonfeature points between two images and then apply hierarchical clusteringmethods to one of them to discriminate between different visual patterns.In order to achieve this last task we introduce a technique to obtain anasymmetric dissimilarity function AX(x; x0) between the nodes X of thenetwork N = (X;AX). For each node, the method weighs the distancebetween each node and the distance with all the other neighbors. Adendrogram is later obtained as the output of the hierarchical clusteringmethod. Finally we show a criteria to select one of the multiple partitionsthat conform the dendrogram.