Word Grouping in Document Images Based on Voronoi Tessellation

Abstract

Abstract. Voronoi tessellation of image elements provides an intuitive and appealing definition of proximity, which has been suggested as an effective tool for the description of relations among the neighboring objects in a digital image. In this paper, a Voronoi tessellation based method is presented for word grouping in document images. The Voronoi neighborhoods are generated from the Voronoi tessellation, with the information about the relations and distances of neighboring connected components, based on which word grouping is carried out. The proposed method has been evaluated on a variety of document images. The experimental results show that it has achieved promising results with a high accuracy, and is robust to various font types, styles, sizes, skew angles, as well as different text orientations.

A contour distance-based approach for multi-oriented and multi-sized character recognition

“…Although there are a few pieces of published work on English stylistic text recognition (Hase et al 2003, Hase et al 2001, Loo & Tan 2002, Lu et al 2004, Sato et al 2000, Tang et al 1991, Uchida et al 2007, Uchida et al 2006, Xie & Kobayashi 1991 but no work is done on the recognition of Indian stylistic text documents. There exists a work (proposed by Pal and Roy (2004)) on Indian stylistic documents but it deals with extraction of individual text lines from Indian stylistic documents.…”

“…In this paper, we propose a scheme towards the recognition of printed multi-oriented and multi-sized Bangla and Devnagari characters and the recognition of multi-oriented and multi-sized characters is done using rotation invariant features. Some pieces of work on English stylistic text recognition are available in the literature (Hase et al 2003, Hase et al 2001, Loo & Tan 2002, Lu et al 2004, Sato et al 2000, Tang et al 1991, Uchida et al 2007, Uchida et al 2006, Xie & Kobayashi 1991. Xie & Kobayashi (1991) proposed a rotation invariant recognition system using the patterns of different angular variation of the component and 97% recognition accuracy is obtained from the 10 digits of English alphabet.…”

Using a Neighbourhood Graph Based on Voronoï Tessellation with DMOS, a Generic Method for Structured Document Recognition

“…Thanks to this information, some authors have proposed various methods to segment documents into words, lines, paragraphs, columns. Generally, like in [4] or [6], the knowledge that is necessary for the analysis is included by learning thresholds like inter-character, inter-word and inter-textline gaps.…”

“…Generally, like in [4] or [6], the knowledge that is necessary for the analysis is included by learning thresholds like inter-character, inter-word and inter-textline gaps. However, this statistic analysis is made on the whole document, which requires the document to be homogeneous enough.…”

Recognition-Based Segmentation of Nom Characters from Body Text Regions of Stele Images Using Area Voronoi Diagram

“…For example, Kise et al [8] and Lu et al [9] used area Voronoi diagram for page segmentation and word grouping in document images respectively using the distance and area ratio between neighboring connected components.…”