Formation and extraction of Chinese character sub-strokes

A novel method of sub-stroke extraction based on the formation law of sub-strokes for Chinese character recognition is presented. From the view of connected pixel segments on the rows (columns) of the image, authors find out that the sub-strokes from a Chinese character image fall into two categories: Ladder sub-stroke and parallel long-sub-stroke. The formation law of these two kinds of sub-strokes is induced from positional relations between connected pixel segments on the rows (columns), and consequently the sub-stroke extraction algorithm is developed. The ladder sub-strokes and parallel long-sub-strokes in the algorithm are extracted one by one from Chinese character image according to their formation law, and then the extracted sub-strokes are merged, thinned and adjusted to get the final results. Experimental results for several fonts of printed Chinese characters and handwritten Chinese characters show that the proposed method is effective and robust.