Detecting Recurring Deformable Objects: An Approximate Graph Matching Method for Detecting Characters in Comics Books

摘要: Graphs are popular data structures used to model pair wise relations between elements from a given collection. In image processing, adjacency graphs often represent the segmented regions. The comparison of such has been largely studied but graph matching strategies essential find, efficiently, similar patterns. this paper, we propose method detect recurring characters in comics books. We would like draw attention reader. term “character” means protagonists story. our approach, each panel is represented with an attributed graph. Then, inexact strategy applied find among set graphs. main idea that same character will be by subgraphs different panels where it appears. two-step process consists node step and edge validation step. Experiments show approach able consequently recurrent book. originality no prior object required characters. algorithm detects, automatically, all corresponding