Computer vision applied to comic book image
Christophe Rigaud and Jean Christophe Burie
University of La Rochelle, France
: J Comput Eng Inf Technol
Abstract
Heritage museums such as the Kyoto International Manga Museum, the International City of Comic books and Images in France (CIBDI) and The Digital Comic Museum (DCM) have already digitized several thousands of comic books that some are now in the public domain. Despite the growing market place of digital comics, few researches have been carried out to take advantage of the added value provided by these new media. Comic book image processing is at the intersection of several research fields within the computer vision domain (e.g. complex background, semi-structured and mixed document images) and combines their difficulties. We review, highlight and illustrate these challenges in order to give a good overview about the last research advances in this field and the current issues. In order to cover the widest possible scope of study, we propose different approaches for comic book image analysis. One aims at recognizing graphical and textual elements in an intuitive way, trying to imitate the human understanding system, from simple to complex visual elements. Simple elements such as panels, balloons and text are recognized first, followed by balloon tails and then the comic character positions. In another approach, we introduces a knowledge-driven system that combines low and high level processing to build a scalable system of automatic comics image understanding. We built an expert system composed by an inference engine and two models, one for comic’s domain and another one for image processing domain, both stored in ontology. This expert system allows a higher level semantic description and consistency. We can then infer the reading order, the semantic of the speech balloons, the relations between speech balloons and speakers, interaction between comic characters etc.
Biography
Email: christophe.rigaud@univ-lr.fr
