摘要

Feature point extraction plays important role in off-line handwritten Chinese character, and the corner point extraction is the most difficult issue. A corner point extraction algorithm based on detecting double angles from macroscopic view is present in this paper. Some candidate pixels are selected with fixed stride from the Chinese character skeleton image, the angles from one candidature to the next are monitored and compared with thresholds to determine whether the candidate pixel is corner point. According to the characteristic of Chinese character, various circumstances are analyzed in this paper, and the experiment on the unconstrained handwritten Chinese character shows that this method is effective and reflects the nature of corner point, the feature points are suitable for describing the characteristics of off-line handwritten Chinese characters.

  • 出版日期2014

全文