Abstract: Calligraphy strokes bear the writing emotions, which distinguish them from printed characters. They are the basic elements for calligraphic character recognition, style classification and writer identification. However due to the complexity of these strokes, there has not been a practical way to extract them yet. This paper presents a new method to extract individual calligraphy strokes. First, one-pixel wide skeletons of a calligraphic character are computed by the corrosion. Then, stroke crawlers walk along these skeletons to obtain the moving trajectory. When meeting a cross, agent crawlers are sent to explore the potential right ways according to the writing rules. After an individual stroke is obtained, its trajectory is coded using 8-directional chain code and used to compute stroke type. According to the stroke type, crossing area and corresponding contours are assigned to different strokes. Finally, all the individual strokes with width of the same character are extracted. Experimental results show that this calligraphy stroke extracting approach works for the seal style, the clerical style, the standard style, and set a foundation for those deformed calligraphy stroke extraction, such as the running style and the cursive style.