Delaunay and Graph Modal-Based Bagua Layout Text Line Extraction for Shui Script

Text line extraction is an important part of document image analysis. It provides significant information for follow-up character recognition and text-based retrieval. By analyzing the layout style and writing features of a document image with radicalized Bagua layout in Jiugong in Shui script, we propose a multi-directional text partition method for Shui Script based on Delaunay triangular mesh and undirected network diagram connected component search technique. Through several steps, including preprocessing, construction of a Delaunay triangular mesh, generation of an undirected network, and generation of multi-directional text, line segmentation of text with Bagua layout of Shui Script is realized. Experiment results reveal that the proposed method has an accuracy rate of almost 90%.