In this paper, we present a hierarchical stroke extraction algorithm based on Stroke Segments and rough clustering, which is the pre-processing step for CJK outline font structure-based compression, and consists of two levels. In the higher level, Stroke Segments are extracted from CJK outline font, in order to remove the intersections and some attachments. Then the strokes are constituted by merging the Stroke Segments, and roughly clustered by the constitution and font knowledge. In the lower level, the strokes which are unsatisfied ones determined by the clustering results are sub-extracted to the satisfied ones according to the constituted Feature Segment. The algorithm can resolve two difficulties in traditional methods. And the experiments reveal that the approach is effective and suitable for multi-fonts.
展开▼
