The old newspaper project

• Main Author: Mao, Junke
• Other Authors: Ling Keck Voon
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/157550
• Institution: Nanyang Technological University

Optical Character Recognition (OCR) is commonly used nowadays for printouts and documents conversion in sociology, communication and education studies. In traditional OCR models, texts are extracted sequentially within the whole page. In the case of newspaper, texts are arranged in columns based on articles with images embedded. As a result, the conversion of text materials with such a complex layout, such as multi-column text, headlines, embedded figures, etc, might impair the outcomes of the OCR results. To improve the efficiency of converting images of newspapers, we built a specialized model for newspaper recognition. The integrated model will perform object segmentation to extract the relevant components in the image, i.e., the headlines, embedded figures, etc, and performs OCR on these components accordingly. The output would be text document logically arranged with headlines, text body in single column, and embedded images appended at the end.