Дослідження методів сегментації зображень документів

Abstract

This paper investigates the methods of document image segmentation with a specific focus on structural layout analysis. The study explores the application of hierarchical representations, namely M-X-Y trees, for the decomposition of document images into logical blocks (text, graphics, tables). The described method demonstrates high efficiency in establishing spatial relationships between segmented elements. The results confirm that utilizing M-X-Y trees significantly improves the accuracy of complex document reconstruction during the optical character recognition (OCR) process.