We present an approach to creating Chinesecalligraphy with particular style from learning author’swritten works automatically. Our first contribution is amethod to represent Chinese character topology via fuzzyrelative position. Our second contribution is an algorithm totake topological features as features fed into the evaluationmodel, which is a decision tree algorithm. Our thirdcontribution is an algorithm of heuristic search in theoptimization glyph phase. To improve the converge speed,we introduce hypothesis testing and the decay function oftransformation amplitude. The experiments demonstrateour algorithm can obtain the similar style Chinesecalligraphy with training samples.
展开▼