Offline Handwritten Chinese Character Segmentation Based On Neighborhood-covering

This paper proposes an offline handwritten Chinese character segmentation algorithm, which works by coarse segmentation and fine segmentation, by statistics and a deep analysis of offline handwritten Chinese character images. After the pretreatment, the offline handwritten Chinese character image is firstly separated into the images of the text lines by the horizontal projection, and the image containing the oblique text lines is dealt with the subsection projection. Secondly, the image of the text lines is divided into the single Chinese character images, and then some combination of the single character images, which is referred to as field, are obtained by the vertical projection. The normal width of a single Chinese character is estimated by the mean of the fields' height, which is got by statistics, and the type of the field is judged, then the coarse segmentation is finished. The fine segmentation works by applying the different segmentation strategies according to the type of the field. As for components, the strategy of combination should be taken. A for the touching characters, the neighborhood-covering algorithm should be taken, so that the non-linear segmentation path will be obtained. The results show that the algorithm proposed in this paper is valid.