Calligraphy Stroke Extraction in CADAL

Abstract

Stroke is the basic unit of Chinese characters. Stroke extraction is the key procedure for calligraphy reconstruction in modern digital libraries, which faces the difficulties caused by deformation and tedious manual work. A novel stroke extraction approach for Chinese calligraphy character is proposed in this paper. Firstly, the corner points on the contours are detected with fast-CPDA, and a half-edge coloring graph is presented to help connecting the contour segments on the same side of a stroke. Then the connected contour segments on opposite sides are matched to form the whole strokes. Finally, the results are refined by rule-based verification. Experiments show that the proposed method can extract basic strokes for simple calligraphy characters automatically and help stroke extraction for complicated calligraphy characters with little manual work.