Scholarlink

<p>Newspapers constitute a rich source of historical and informational content, and extensive efforts are underway to preserve them through digitization. Although scanning newspapers as images preserves their visual structure, it limits text-based retrieval and analysis. Optical Character Recognition (OCR) addresses this limitation by converting scanned images into searchable, machine-readable text; however, its performance heavily depends on accurate document segmentation, particularly for complex newspaper layouts. This paper presents a robust segmentation framework for Devanagari-script newspaper articles that effectively handles both text-only and mixed-content documents. The proposed approach first separates graphical and non-textual components to enable accurate extraction of textual regions. Subsequently, it performs hierarchical segmentation of headlines, body text, columns, text lines, words, and individual characters. The framework is designed to address challenges such as multi-column layouts, varying font sizes, degraded print quality, and script-specific characteristics of Devanagari. Experimental evaluation conducted on 50 Hindi newspaper articles demonstrates segmentation accuracies of 87.5% for text–graphic separation, 85.0% for headline and column segmentation, 93.27% for text line segmentation, and 98.61% for character segmentation. These results confirm the effectiveness and robustness of the proposed approach, making it a strong foundation for developing reliable OCR systems for Devanagari newspapers.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Segmentation of Devanagari Newspaper Articles for Enhanced OCR Performance

  • 期刊论文

  • 发表日期：

    2026年4月9日

查看全文
  • Ahmed Alkhayyat,
  • S. Narayanasamy,
  • Ibrahim Oteir,
  • Deeksha Verma,
  • K. N. Raja Praveen,
  • Sarbeswara Hota,
  • Rajesh Singh,
  • Saurabh Namdev

摘要

Newspapers constitute a rich source of historical and informational content, and extensive efforts are underway to preserve them through digitization. Although scanning newspapers as images preserves their visual structure, it limits text-based retrieval and analysis. Optical Character Recognition (OCR) addresses this limitation by converting scanned images into searchable, machine-readable text; however, its performance heavily depends on accurate document segmentation, particularly for complex newspaper layouts. This paper presents a robust segmentation framework for Devanagari-script newspaper articles that effectively handles both text-only and mixed-content documents. The proposed approach first separates graphical and non-textual components to enable accurate extraction of textual regions. Subsequently, it performs hierarchical segmentation of headlines, body text, columns, text lines, words, and individual characters. The framework is designed to address challenges such as multi-column layouts, varying font sizes, degraded print quality, and script-specific characteristics of Devanagari. Experimental evaluation conducted on 50 Hindi newspaper articles demonstrates segmentation accuracies of 87.5% for text–graphic separation, 85.0% for headline and column segmentation, 93.27% for text line segmentation, and 98.61% for character segmentation. These results confirm the effectiveness and robustness of the proposed approach, making it a strong foundation for developing reliable OCR systems for Devanagari newspapers.