Similarity Shape Based on Skeleton Graph Matching

Using graphs to match two feature sets through embedded high-order relations points has many possible applications in criminal justice, security, and high technology. In this paper, we analyze the method of using the random walk framework to establish correspondence between two skeleton graphs and find out matching points between two shapes. The graphs are matched using a skeleton graph with the descriptors of the relationship between the two edges of the end-nodes ranked on an association graph. Through adopting individual jumps with a reweighting scheme, the new proposed approach effectively reflects the one-to-one matching constraints during the random walk process. Experiments on several benchmark data sets show that the proposed approach clearly outperforms existing algorithms, especially in the presence of noise and outliers.