Relational indexing of vectorial primitives for symbol spotting in line-drawing images

Abstract

This paper presents a symbol spotting approach for indexing by content a database of line-drawing images. As line-drawings are digital-born documents designed by vectorial softwares, instead of using a pixel-based approach, we present a spotting method based on vector primitives. Graphical symbols are represented by a set of vectorial primitives which are described by an offthe-shelf shape descriptor. A relational indexing strategy aims to retrieve symbol locations into the target documents by using a combined numerical-relational description of 2D structures. The zones which are likely to contain the queried symbol are validated by a Hough-like voting scheme. In addition, a performance evaluation framework for symbol spotting in graphical documents is proposed. The presented methodology has been evaluated with a benchmarking set of architectural documents achieving good performance results.

Graphical Symbol Recognition

“…To compute spatial relations, a common angle-based theory based on star-calculus is shown in Fig 13. In document image analysis, precise relations are required to analyze/understand scanned architectural and color map documents (270), recognize models (271), and to define efficient retrieval methods (272)(273)(274)(275). Similarly, authors have shown the usefulness of relational indexing of vectorial primitives for line drawing images (276). Considering the problem of symbol localization in real documents, composed of individual parts and constrained by spatial relations, as said before, global signal-based descriptors cannot be applied since they are, unfortunately, primarily designed for applications where symbols are isolated.…”

Integrating vocabulary clustering with spatial relations for symbol recognition

“…• F E : E → R E is a function assigning labels to the edges where R represents spatial relations of the edge E as developed in [Santosh et al, 2011b]. Effects of spatial relations on recognition performance have already been examined comprehensively for scene understanding [Biederman, 1972;Bar and Ullman, 1993], document analysis and recognition tasks [Xiaogang et al, 2004;Pham and Smeulders, 2006;Rusiñol et al, 2010;Santosh et al, 2011b]. Based on [Freeman, 1975], spatial relations can be 1) topological [Egenhofer and Herring, 1991] such as disconnected, externally connected, overlap, and contain or inside; 2) directional [Bloch, 1999;Matsakis and Wendling, 1999;Wang and Keller, 1999] such as north and south; and 3) metric such as near or far.…”

“…Among the existing state-of-the-art, structural approaches provide powerful representations, conveying how parts are connected to each other. However, relations [Rusiñol et al, 2010;Santosh et al, 2011b] do not exploit shape information as shape descriptors do. On the other hand, especially when symbols are not in simple linear and isolated form, but are composed of many elements, sometimes exhibiting subtle differences, global signal-based shape descriptors cannot provide optimal retrieval performance.…”

Reconnaissance de symboles à partir d'une signature structurelle flexible et d'un classifieur de type treillis de Galois

“…Cet article traite du problème de reconnaissance de symboles. Dans la littérature, très abondante dans ce domaine (par exemple Tombre et al (2003), Dosch et al (2004), Luqman et al (2009) ou Rusiñol et al (2009a)), la reconnaissance de symboles est souvent définie, comme la plupart des tâches de reconnaissance des formes, comme un processus en deux étapes : l'extraction de signatures et la classification.…”