Computer Vision Applied to Comic Book Images
Christophe Rigaud, Jean-Christophe Burie · 2018
This chapter addresses the problem of detecting visual and textual entities using pattern recognition algorithms, and understanding comics&s; content by automatically improving the consistency of the extracted elements. It overviews the existing research on holistic understanding of documents and the relevant computer science literature on comic book image analysis. One of the original goals of image document analysis was to fully understand the content of any given image. Images in comics are mixed content documents that are processed differently depending on the type of content that we are interested in. The techniques involved can vary a lot depending on whether we focus on panels, balloons, text, or processing comic characters. The chapter provides the low- and high-level system algorithms and their interactions that compose the proposed framework. The shape, size, and position of comic book images, color, and drawing style all carry strong semantic information. Inspiration may come from the increased interest that psychologists and linguists shows in comic books.