A Stroke-Density Based Double Elastic Meshing Feature Extraction Method for Chinese Handwritten Character Recognition

摘要

Stroke-density based elastic meshing method can not only absorb the deformations of stroke in different handwritings but also avoid the non-uniform width of strokes which caused by using a nonlinear normalization method. In view of existing situation that the stroke density functions do not consider the distribution of strokes in diagonal direction, the stroke-density based vertical-horizontal elastic meshing method could not effectively extract the feature of left-falling and right-falling stroke in a Chinese handwritten character image, we propose a new stroke density definition of a diagonal direction and combine it with diagonal elastic meshing technique to constitute meshes. Then, combining the vertical-horizontal elastic meshes with the diagonal's, we get the stroke-density based double elastic meshing approach. The experimental results have verified the effectiveness of this method.