Fig. 1. Partial illustration of lines in ancient Tibetan texts. (a) Baseline and syllable points; (b) slanted text line; (c) distorted text line; (d) overlapping and adhesion of text lines

Fig. 2. Flow chart of method

Fig. 3. Tibetan historical document image

Fig. 4. Syllable point detection of Tibetan historical document. (a) Binary map after layout processing (black background area is local binary map); (b) syllable point image

Fig. 5. Projection images. (a) Horizontal projection of syllable point image; (b) vertical projection of binary original image

Fig. 6. Formation of pseudo-text connected regions. (a) Image of text core area; (b) image of pseudo-text connected region

Fig. 7. Schematic of pixel point expansion growth process. (a) Pixel point without expansion; (b) pixel point during expansion; (c) pixel point after expansion

Fig. 8. Line acquisition of pseudo text. (a) Expansion growth result; (b) pseudo-text lines

Fig. 9. Local diagram in attribution process of broken strokes. (a) Text lines to be split; (b) pseudo-text lines; (c) complete split lines

Fig. 10. Final segmentation result of local image

Fig. 11. Text lines and their segmentation results. (a)(c)(e) Original images; (b)(d)(f) segmentation results

Fig. 12. Segmentation effects of overlapped, adhered and broken strokes. (a)(b) Overlapped strokes; (c)(d) adhered strokes; (e)(f) broken strokes

Fig. 13. Wrong segmentation and corresponding correct forms. (a)(c)(e)Wrong segmentation; (b)(d)(f)corresponding correct forms