基于字体特征与多尺度PatchGAN的中文字体风格转换研究

Chinese font style transfer research based on font features and multi-scale patch generative adversarial network

针对现有中文字体风格转换方法生成字符图像质量低以及生成字体图像与目标字体图像风格不一致的问题,提出基于字体特征与多尺度patch生成对抗网络的中文字体风格转换方法.首先,根据字体特点设计两个特征提取网络,分别提取字体风格特征和字符内容特征;然后,将两个特征输入生成器,利用字体风格特征约束生成字符图像的风格,字符内容特征约束生成字符图像的字形;最后,将生成字符图像输入到多尺度patch判别器中,对生成结果的多尺度图像块判断真假.实验结果表明,所提方法有效提升了生成字符图像的质量以及与目标字体的风格一致性.

Aiming at the problem of low quality of the generated character images and the style inconsistency between the generated font images and the target font images in existing font style transfer methods,a Chinese font style transfer method based on font features and multi-scale patch generation adversation network was proposed.Two feature extraction networks were designed according to the characteristics of fonts to extract font style features and character content features respectively.The two features were inputted into the generator,then the font style feature was used to constrain the style of the generated character images,and the character content feature was used to constrain the glyph of the generated character images.Finally,the generated character images were inputted into a multi-scale patch discriminator to judge the true or false of multi-scale patches.Experimental results shown that the proposed method effectively improved the quality of generated character images and the style consistency with the target font.