Abstract: Traditional making technology of Chinese character database requires a large number of manual operations in the character-formation results. In order to solve this defect, a method of achieving quality evaluation for the character-formation results is proposed by using attributes of Chinese characters. First, this paper extracts the attributes which could reflect the standard writing rules of Chinese characters through image processing technology, to quantify the evaluation criteria of Chinese characters, and then, models the font quality evaluation making use of the improved three-layer BP neural network, which would provide basis for adjustment of the font and reduce the human input in the process of character making.

Key words:  attributes, quality evaluation, image processing, quantization, BP neural network