摘要
为了解决传统彩色铅笔画算法生成结果单一的问题,提出了一种基于卷积神经网络(CNN)生成彩色铅笔画的算法.采用分数阶微分获取原始图像轮廓信息,用卷积神经网络获取艺术家手绘铅笔画风格,利用直方图匹配获取与手绘铅笔画相似的色调,并使用L-BFGS优化算法来合成具有铅笔画效果的图像.该算法能够生成具有不同风格的彩色铅笔画图像.实验结果表明,该算法生成的图像保留了更多原始图像的细节信息,风格更加灵活多样.
In order to optimize the single generation result of traditional color pencil drawing algorithms,a convolution neural network(CNN)based color pencil drawing generation method is presented.Fractional differentiation is employed to obtain original image contour information,CNN can obtain pencil drawing style,and histogram matching can obtain similar tones.Meanwhile,L-BFGS algorithm is used to synthesize pencil drawing image.This can generate color pencil drawing images of different styles.Experiments show that the images generated can retain more original image detail information,and feature with more flexible and diverse styles.
作者
王小玉
胡鑫豪
韩昌林
WANG Xiao-yu;HU Xin-hao;HAN Chang-lin(School of Computer Science and Technology,Harbin University of Science and Technology,Harbin 150080,China)
出处
《北京邮电大学学报》
EI
CAS
CSCD
北大核心
2020年第1期129-134,共6页
Journal of Beijing University of Posts and Telecommunications
基金
国家自然科学基金项目(60572153,60972127).
