Perceptually Extracting Multi-Oriented Text Lines of Online Handwriting

This paper presents an approach to perceptually extract multi-oriented text lines from online handwriting. The approach adopts a bottom-up strategy. By the approach, strokes are first merged into blocks analogous to words, then a link model is constructed upon these blocks. Finally, the optimal text line layout is found out by a branch-bound search on the link model. The experimental results show that the approach is effective and also available to extract curved text lines.