Effective radical segmentation of offline handwritten Chinese characters towards constructing personal handwritten fonts

Effective radical segmentation of handwritten Chinese characters can greatly facilitate the subsequent character processing tasks, such as Chinese handwriting recognition/identification and the generation of Chinese handwritten fonts. In this paper, a popular snake model is enhanced by considering the guided image force and optimized by Genetic Algorithm, such that it achieves a significant improvement in terms of both accuracy and efficiency when applied to segment the radicals in handwritten Chinese characters. The proposed radical segmentation approach consists of three stages: constructing guide information, Genetic Algorithm optimization and post-embellishment. Testing results show that the proposed approach can effectively decompose radicals with overlaps and connections from handwritten Chinese characters with various layout structures. The segmentation accuracy reaches 94.91% for complicated samples with overlapped and connected radicals and the segmentation speed is 0.05 second per character. For demonstrating the advantages of the approach, radicals extracted from the user input samples are reused to construct personal Chinese handwritten font library. Experiments show that the constructed characters well maintain the handwriting style of the user and have good enough performance. In this way, the user only needs to write a small number of samples for obtaining his/her own handwritten font library. This method greatly reduces the cost of existing solutions and makes it much easier for people to use computers to write letters/e-mails, diaries/blogs, even magazines/books in their own handwriting.