用于字体识别的Gabor滤波角度GA优化方法

The Orientation Optimization in Gabor Filter Used in Chinese Font Recognition Based on GA

Gabor滤波角度对字体识别结果有重要影响,由于字体纹理与自然纹理的不同,现有的Gabor滤波器角度参数不适于提取字体纹理的有效特征。基于字体纹理的多变性,该文提出使用遗传算法通过对字体纹理的学习优化滤波角度参数,使之能够适应字体纹理的特点,以提高识别率。通过对常用的4种字体899块字体纹理样本的测试表明:遗传算法能够找到适合字体识别的角度参数,使用新的角度参数减少了识别时间,提高了字体识别率。

The selection of orientation in Gabor filters has a strong effect on the result of font recognition.Due to the difference between natural texture and font texture,the current orientation parameters of Gabor filter are unfit to extract the effective font texture features.Based on the variability of font textures,Genetic Algorithm is adopted in this paper with which the orientation values can be adjusted to the font texture's character by studying the known textures in the hope of improving the recognition rate.At the end,experiments are carried out by using899font texture samples of4fonts used in common,the results show that fitting orientation values can be found by Genetic Algorithm,and using those new values time can be saved and font recognition rate can be improved.