基于脉冲耦合神经网络的汉字纹理特征提取

Pulse Coupled Neural Network- Based Chinese Characters Texture Feature Extraction

论文运用改进的脉冲耦合神经网络(PCNN)简化模型结合梯度向量的方法对汉字图像进行纹理特征提取。首先对汉字图像求出梯度向量得到梯度图像,然后利用PCNN的脉冲并行高速传播特性对梯度图像进行迭代点火,每次点火后的二值图像进行概率统计,全部迭代次数的统计向量作为提取的纹理特征。仿真结果表明统计向量作为纹理特征的有效性,同时验证了该方法具有运算速度快、旋转不变性和尺度不变性的优点,为汉字复原提供了研究的基础条件。

A method for Chinese characters image texture feature extraction based on improved Pulse Coupled Neural Network （PC- NN） integrated with gradient vector is proposed in this paper. Producing the gradient image of original image by computing gradient vector firstly, then an high - speed parallel spreading characteristic of PCNN is employed for iterative firing. Finally, probability statis- tics of binary image fired each time is extracted as texture feature. The simulation shows that this method has the advantages of high speed, rotation invariance and scale invariance, providing the basis for restoration of Chinese characters.