Stroke extraction of Chinese characters plays an important role in Chinese character information processing such as character recognition, document analysis, document compression and storage, font automation and so on. By analyzing the structure of Chinese characters deeply, this paper developed a novel method to extract strokes of Chinese characters directly from the original character pattern image. Two theorems, eight rules and an algorithm for stroke extraction of Chinese characters are presented. This method can overcome the difficulties encountered in disposing the intersection or connection of different strokes, and can eliminate noises successfully. Our experiments have shown that this method can extract strokes both accurately and efficiently.