二值字符图案细化的一种算法——神经网络算法

AN ALGORITHM FOR THINNING BINARY CHARACTER PATTERNS -A NEURAL NETWORK ALGORITHM

为了识别二值图案，常常对它先进行细化处理，本文提出了基于神经网络的细化处理算法．首先，讨论了图案细化所用的ＢＰ网络结构，确定其结构为８－１０－４；然后，利用ＢＰ神经网络的分类识别特性来识别图案的边缘点，从而达到是否删除它的目的；最后，描述了字符的ＢＰ网络细化算法，并给出了图案细化的结果，得到的结果扭曲程度很小，并且比较光滑．

Thinning of Binary character patterns is usually carried out for recognizing it. The thinning algorithm based on neural network is proposed. Firstly, structure of BP network for the thinning is discussed and determined as 8 10 4; then edge point of the pattern is recognized out by using classification and recognition characteristics of BP network, so it is decided that the point should be deleted or not; finally,the thinning algorithm is detailed and the skeleton of the character pattern is given. Experiment shows that the thinning skeleton is less sprained and much smoother than one thinned by other papers before.