Robust Frame and Text Extraction from Comic Books

Abstract

Abstract. Comic books constitute an important heritage in many countries. Nowadays, digitization allows to search directly from content instead of metadata only (e.g. album title or author name). Few studies have been done in this direction. Only frame and speech balloon extraction have been experimented in the case of simple page structure. In fact, the page structure depends on the author which is why many different structures and drawings exist. Despite the differences, drawings have a common characteristic because of design process: they are all surrounded by a black line. In this paper, we propose to rely on this particularity of comic books to automatically extract frame and text using a connectedcomponent labeling analysis. The approach is compared with some existing methods found in the literature and results are presented.

Digital Comics Image Indexing Based on Deep Learning

“…These methods do not consider empty area [15] and border-free panels. These issues have been corrected by connected component labeling approaches, but these approaches are sensitive to regions that sometimes connect several panels, which potentially increases the detection error rate [20,21]. Another approach based on morphological analysis and region growing can remove such connecting elements but at the same time create new holes in the panel border [22].…”

eBDtheque: A Representative Database of Comics

“…Indeed, the efficiency of an algorithm will depend on its ability to deal with position, shape and style of the elements, as well as their graphical relationship, it is aiming to recognize. For example, for panel detection there are several methods either based on corners and edges [16], connected components [6] or straight lines [2] but they all fail for certain comics' styles. Considering that point, several comic books authors with their own specific style have been asked to provide material for the construction of this ground truth database.…”

“…The two main bottlenecks to these goals achievements are the image segmentation and the semantic retrieval. The first problem is addressed in numerous papers with a special focus on panel extraction [1], [2], speech balloons [3], [4] and text recognition [5], [6]. But we are also witnessing a growing interest for the latter.…”

Automatic Text Localisation in Scanned Comic Books

“…As the morphological operations are performed with a fixed mask size, the method is orientation and resolution dependent. (Rigaud et al, 2012) make use of "the median value of the border page pixels" to binarize the image, extract CC and then classify them into "noise", "text" or "frame" (based on CC heights). This method assumes that the page always contains text and that the text background colour is similar to the paper background.…”