A Robust Segmentation Method for Chinese Inscription Images

Abstract

Inscriptions are important carriers of Chinese calligraphy which has high calligraphic, artistic and cultural value. Segmentation of Chinese inscription images play a fundamental role in processing of Chinese character images. Traditional projection segmentation methods can be used for Chinese inscription image segmentation because Chinese characters in most inscription images arrange regularly. However, in an actual segmentation process, the slight offset of the glyph position or a crossing of strokes after projection caused projection segmentation method failure. In order to solve this problem, a Chinese inscription image segmentation method based on clustering algorithm was provided. Firstly, preprocessed the image, and then looked up the contour of the inscription images to exclude the obvious abnormal part of the contour size; Secondly, filled in the remaining contours to obtain a clustered sample set; Thirdly, used the DBSCAN clustering algorithm to cluster the sample set to generate several clusters which each of those clusters represented a Chinese character image; Finally, for clusters that did not satisfy the outline range of Chinese characters, adjusted the minimum neighborhood until the aspect ratio of the circumscribed rectangle of the cluster was within the range of the aspect ratio of Chinese characters. We conducted character segmentation experiments on sample inscriptions. Results showed that out method not only segmented and processed the regularly arranged inscriptions images, but also worked efficiently for irregularly arranged inscriptions images.