A texture-based pixel labeling approach for historical books

Abstract

Over the last few years, there has been tremendous growth in the automatic processing of digitized historical documents. In fact, finding reliable systems for the interpretation of ancient documents has been a topic of major interest for many libraries and the prime issue of research in the document analysis community. One important challenge is to refine well-known approaches based on strong a priori knowledge (e.g. the document image content, layout, typography, font size and type, scanning resolution, image size, etc.). Nevertheless, a texture analysis approach has consistently been chosen to segment a page layout when information is lacking on document structure and content. Thus, in this article a framework is proposed to investigate the use of texture as a tool for automatically determining homogeneous regions in a digitized historical book and segmenting its contents by extracting and analyzing texture features independently of the layout of the pages. The proposed framework is parameter-free and applicable to a large variety of ancient of books. It does not assume a priori information regarding document image content and structure. It consists of two phases: a texture-based feature extraction step and unsupervised clustering and labeling task based on the consensus clustering, hierarchical ascendant classification, and nearest neighbor search algorithms. The novelty of this work lies in the clustering of extracted texture descriptors to find automatically homogeneous regions, i.e. graphic and textual regions, by using the clustering approach on an entire book instead of processing each page individually. Our framework has been evaluated on a large variety of historical books and achieved promising results.

Texture feature benchmarking and evaluation for historical document image analysis

“…Tamura -Zone classification [41] -Content segmentation [55] LBP -Text localization [9] -Pixel classification [15,16] -Printed script identification [26] -Arabic font recognition [57] GLRLM -Pixel classification [36] -Zone classification [41] -Text line, word and character segmentation [58] -Handwritten annotation discrimination [66] -Content segmentation [73] Auto-correlation -Geometric layout analysis [18] -Handwriting classification [25] -Text recognition [29] -Layout analysis and content enrichment [31] -Pixel-labeling for historical books [40,51] GLCM -Content segmentation and classification [48,54] -Lettrine retrieval [71] Gabor -Detection of main text area from side-notes [5] -Pixel classification [16] -Text region segmentation [21] -Text/drawing separation [25] -Text detection [40] Wavelet -Text localization [44] -Content classification [43] Due to the large diversity of texture-based methods, we conclude that there is a critical need to explore and compare various aspects of texture features by using a classical texture-based pixel-labeling scheme in order to assist HDIA and clarify a number of issues. Which texture features are firstly well suited for segmenting graphical contents from textual ones, discriminating text in a variety of situations of different fonts and scales and separating different types of graphics?…”

“…Indeed, without hypothesis on either the DI layout or content, the choice of numerous appropriate thresholds and parameters is a very difficult task. In addition, the pre-defined parameters used when extracting and analyzing the texture features in our study are set up based on work published in the literature (Tamura [41,55], local binary patterns [9], gray-level runlength matrix [73], auto-correlation function [40,51], graylevel co-occurrence matrix [11], Gabor filters [38], 3-level Haar wavelet transform, 3-level wavelet transform using 3tap Daubechies filter and 3-level wavelet transform using 4tap Daubechies filter [44]). In addition, they are the most classic and common ones in the literature.…”

“…The auto-correlation features which have been investigated in this article, are summarized in Table 5 [40,51].…”

“…The silhouette width (SW ) which is considered as an internal or unsupervised per-pixel accuracy metric, measures the level of compactness and separation by analyzing the distribution of the observations into clusters. For assessing the quality of segmenting historical document images into homogeneous or similar regions defined by similar texture indices, Mehri et al [51] defined an external or supervised evaluation metric, called purity per-block (PPB). The PPB computes the homogeneity rate of regions by assessing the matching regions between the defined ground truth and the obtained pixel-labeling results.…”

Text Extraction in Document Images: Highlight on Using Corner Points

“…Finally, other methods are based on pixel classification: with clustering [37]; after a characterization at pixel level by textures [33] [34]; using Markov Random Field [35]; or statistical methods [36]. In [8], authors have used inpainting.…”