摘要
目的为了更好地实现包装设计中符合用户美学感知需求的动漫图像元素,生成美感更高的彩色线稿图像,提出一种基于美学质量评价的条件生成对抗网络线稿图像彩色化方法(Conditional Image Colorization with Image Aesthetics GANs,IM-GAN)。方法采用Mish函数作为生成模型激活函数,并使用美学质量评价优化生成模型损失函数,实现线稿图像的自动彩色化任务。结果使用图像美学质量评分作为客观评价指标,观察者打分作为主观评价指标,对算法进行评价。IM-GAN生成的彩色图像具有更高美学质量评分和主观打分。结论文中方法能够完成线稿图像自动彩色化任务,在包装设计动漫形象应用方面具有一定的参考和使用价值。
In order to better realize the animation image elements in the packaging design that meet the users'aesthetic perception needs,generate color line draft images with rising aesthetics,a colorization method based on the conditional generation adversarial networks of the aesthetic quality evaluation against the network linear draft image is proposed.The Mish function is used as the activation function of the generation model,and the loss function of the generation model is optimized by the aesthetic quality evaluation to realize the automatic coloring task of the line draft image.The algorithm is evaluated by using image aesthetics quality score as objective evaluation indicators and observer scores as subjective evaluation indicators.The color images generated by IM-GAN have higher aesthetic quality scores and subjective scoring.The results show that the method in this paper can complete the task of automatic colorization of linear images,and has certain reference and use value in the application of animation images in packaging design.
作者
王晓红
刘博伟
谌鹏
WANG Xiao-hong;LIU Bo-wei;CHEN Peng(School of Communication and Art Design,University of Shanghai for Science and Technology,Shanghai 200000,China)
出处
《包装工程》
CAS
北大核心
2021年第15期292-298,共7页
Packaging Engineering
关键词
条件生成对抗网络
图像美学质量评价
线稿图像彩色化
美学质量评分
Conditional generative adversarial networks
image aesthetics quality evaluation
line art image colorization
aesthetic quality score
